ࡱ> ]^_^qym } C E@ jbjbj 0qHT#4(d$4(4(4((-$w@5,lF"FFFG" TX$eRQ wGG w wZZFFe@ w ZtFF wlqVhmv jF45 %G4( wr|(wjx_$_j,2!|ZZx_dZR cjToZZZ$D=,b$PJ"$a0PJAmalgam: A machine-learned generation module Michael Gamon, Eric Ringger, and Simon Corston-Oliver {mgamon, ringger, simonco}@microsoft.com 11 June 2002 Technical Report MSR-TR-2002-57 Microsoft Research One Microsoft Way Redmond WA 98052 USA  TOC \o "2-3" \h \z \t "Heading 1,1"  HYPERLINK \l "_Toc11647534" 1 Introduction  PAGEREF _Toc11647534 \h 1  HYPERLINK \l "_Toc11647535" 2 Prior work in sentence realization  PAGEREF _Toc11647535 \h 2  HYPERLINK \l "_Toc11647536" 3 Properties of German  PAGEREF _Toc11647536 \h 5  HYPERLINK \l "_Toc11647537" 3.1 The Position of the Verb in German  PAGEREF _Toc11647537 \h 5  HYPERLINK \l "_Toc11647538" 3.2 Separable Prefixes  PAGEREF _Toc11647538 \h 8  HYPERLINK \l "_Toc11647539" 3.3 Morphological Case  PAGEREF _Toc11647539 \h 8  HYPERLINK \l "_Toc11647540" 3.4 Constituent Order  PAGEREF _Toc11647540 \h 9  HYPERLINK \l "_Toc11647541" 3.5 Extraposition of Clauses  PAGEREF _Toc11647541 \h 10  HYPERLINK \l "_Toc11647542" 4 The Nlpwin system  PAGEREF _Toc11647542 \h 11  HYPERLINK \l "_Toc11647543" 4.1 The syntactic analysis: Sketch and Portrait  PAGEREF _Toc11647543 \h 11  HYPERLINK \l "_Toc11647544" 4.2 The semantic representation: Logical Form  PAGEREF _Toc11647544 \h 12  HYPERLINK \l "_Toc11647545" 5 The procedural flow  PAGEREF _Toc11647545 \h 13  HYPERLINK \l "_Toc11647546" 5.1 The Major Stages of the Amalgam Pipeline  PAGEREF _Toc11647546 \h 13  HYPERLINK \l "_Toc11647547" 5.1.1 Pre-Processing  PAGEREF _Toc11647547 \h 15  HYPERLINK \l "_Toc11647548" 5.1.2 Flesh-out  PAGEREF _Toc11647548 \h 16  HYPERLINK \l "_Toc11647549" 5.1.3 Conversion to basic tree  PAGEREF _Toc11647549 \h 17  HYPERLINK \l "_Toc11647550" 5.1.4 Global movement  PAGEREF _Toc11647550 \h 18  HYPERLINK \l "_Toc11647551" 5.1.5 Intra constituent ordering  PAGEREF _Toc11647551 \h 19  HYPERLINK \l "_Toc11647552" 5.1.6 Surface cleanup  PAGEREF _Toc11647552 \h 20  HYPERLINK \l "_Toc11647553" 5.1.7 Punctuation  PAGEREF _Toc11647553 \h 21  HYPERLINK \l "_Toc11647554" 5.1.8 Inflectional generation  PAGEREF _Toc11647554 \h 21  HYPERLINK \l "_Toc11647555" 5.2 A Detailed Flowchart of the Amalgam Pipeline  PAGEREF _Toc11647555 \h 23  HYPERLINK \l "_Toc11647556" 6 The rule-based operations in Amalgam  PAGEREF _Toc11647556 \h 26  HYPERLINK \l "_Toc11647557" 6.1 Degraphing  PAGEREF _Toc11647557 \h 26  HYPERLINK \l "_Toc11647558" 6.2 Miscellaneous rule-based operations  PAGEREF _Toc11647558 \h 28  HYPERLINK \l "_Toc11647559" 6.2.1 Creation of lexnodes  PAGEREF _Toc11647559 \h 28  HYPERLINK \l "_Toc11647560" 6.2.2 Simplification of compounds  PAGEREF _Toc11647560 \h 29  HYPERLINK \l "_Toc11647561" 6.2.3 Contracting PPs  PAGEREF _Toc11647561 \h 29  HYPERLINK \l "_Toc11647562" 6.2.4 Insertion of relative pronouns  PAGEREF _Toc11647562 \h 29  HYPERLINK \l "_Toc11647563" 6.2.5 Insertion of reflexive pronouns  PAGEREF _Toc11647563 \h 29  HYPERLINK \l "_Toc11647564" 6.2.6 Insertion of wie  PAGEREF _Toc11647564 \h 30  HYPERLINK \l "_Toc11647565" 6.2.7 Converting the fleshed-out LF to a basic tree  PAGEREF _Toc11647565 \h 30  HYPERLINK \l "_Toc11647566" 6.2.8 The splitting of separable prefixes  PAGEREF _Toc11647566 \h 30  HYPERLINK \l "_Toc11647567" 6.2.9 Introduction of coordination  PAGEREF _Toc11647567 \h 30  HYPERLINK \l "_Toc11647568" 6.2.10 Rule-based movement operations  PAGEREF _Toc11647568 \h 30  HYPERLINK \l "_Toc11647569" 6.2.11 Placement of inflectional features on verbs  PAGEREF _Toc11647569 \h 31  HYPERLINK \l "_Toc11647570" 6.2.12 Fixing up surface order  PAGEREF _Toc11647570 \h 31  HYPERLINK \l "_Toc11647571" 6.2.13 Compound generation  PAGEREF _Toc11647571 \h 31  HYPERLINK \l "_Toc11647572" 6.3 Inflectional generation  PAGEREF _Toc11647572 \h 32  HYPERLINK \l "_Toc11647573" 7 The machine-learned components of Amalgam  PAGEREF _Toc11647573 \h 32  HYPERLINK \l "_Toc11647574" 7.1 Decision Tree Classifiers  PAGEREF _Toc11647574 \h 32  HYPERLINK \l "_Toc11647575" 7.1.1 Syntactic labeling  PAGEREF _Toc11647575 \h 35  HYPERLINK \l "_Toc11647576" 7.1.2 Determiner insertion  PAGEREF _Toc11647576 \h 37  HYPERLINK \l "_Toc11647577" 7.1.3 Auxiliary insertion  PAGEREF _Toc11647577 \h 38  HYPERLINK \l "_Toc11647578" 7.1.4 Preposition insertion  PAGEREF _Toc11647578 \h 39  HYPERLINK \l "_Toc11647579" 7.1.5 Insertion of infinitival marker  PAGEREF _Toc11647579 \h 40  HYPERLINK \l "_Toc11647580" 7.1.6 Negation insertion  PAGEREF _Toc11647580 \h 40  HYPERLINK \l "_Toc11647581" 7.1.7 Insertion of subordinating conjunctions  PAGEREF _Toc11647581 \h 42  HYPERLINK \l "_Toc11647582" 7.1.8 Insertion of expletive subjects  PAGEREF _Toc11647582 \h 43  HYPERLINK \l "_Toc11647583" 7.1.9 Assignment of probabilities for the spellout of NPs  PAGEREF _Toc11647583 \h 44  HYPERLINK \l "_Toc11647584" 7.1.10 Assignment of Case  PAGEREF _Toc11647584 \h 45  HYPERLINK \l "_Toc11647585" 7.1.11 Assignment of verb position features  PAGEREF _Toc11647585 \h 47  HYPERLINK \l "_Toc11647586" 7.1.12 Inversion of dominance  PAGEREF _Toc11647586 \h 49  HYPERLINK \l "_Toc11647587" 7.1.13 Extraposition  PAGEREF _Toc11647587 \h 50  HYPERLINK \l "_Toc11647588" 7.1.14 Realization of determiners  PAGEREF _Toc11647588 \h 52  HYPERLINK \l "_Toc11647589" 7.1.15 Realization of relative pronouns  PAGEREF _Toc11647589 \h 54  HYPERLINK \l "_Toc11647590" 7.1.16 Syntactic aggregation  PAGEREF _Toc11647590 \h 56  HYPERLINK \l "_Toc11647591" 7.1.17 Punctuation  PAGEREF _Toc11647591 \h 58  HYPERLINK \l "_Toc11647592" 7.2 The Order Model  PAGEREF _Toc11647592 \h 60  HYPERLINK \l "_Toc11647593" 7.2.1 Motivation  PAGEREF _Toc11647593 \h 60  HYPERLINK \l "_Toc11647594" 7.2.2 Model and Features  PAGEREF _Toc11647594 \h 61  HYPERLINK \l "_Toc11647595" 7.2.3 Search and Complexity  PAGEREF _Toc11647595 \h 63  HYPERLINK \l "_Toc11647596" 8 Performance  PAGEREF _Toc11647596 \h 63  HYPERLINK \l "_Toc11647597" 9 Evaluation  PAGEREF _Toc11647597 \h 63  HYPERLINK \l "_Toc11647598" 10 Using Amalgam in Machine Translation: First Results  PAGEREF _Toc11647598 \h 64  HYPERLINK \l "_Toc11647599" 11 Conclusion  PAGEREF _Toc11647599 \h 65  HYPERLINK \l "_Toc11647600" 12 Acknowledgements  PAGEREF _Toc11647600 \h 66  Abstract Amalgam is a novel system for sentence realization during natural language generation. Amalgam takes as input a logical form graph, which it transforms through a series of modules involving machine-learned and knowledge-engineered sub-modules into a syntactic representation from which an output sentence is read. Amalgam constrains the search for a fluent sentence realization by following a linguistically informed approach that includes such component steps as raising, labeling of phrasal projections, extraposition of relative clauses, and ordering of elements within a constituent. In this technical report we describe the architecture of Amalgam based on a complete implementation that generates German sentences. We describe several linguistic phenomena, such as relative clause extraposition, that must be handled in order to successfully generate German. Introduction We describe the architecture of a novel sentence realization component, Amalgam (A Machine-Learned Generation Module). Amalgam is a module within the German NLPWIN system at Microsoft Research. The need for a sentence realization module arose in the context of on-going research into machine translation (Richardson et al. 2001a, Richardson et al. 2001b). Sentences in a source language are analyzed to logical forms. These logical forms are transferred to the target language, and must then be realized as fluent sentences. For some target languages we already had mature, high quality knowledge-engineered sentence realization modules (Aikawa et al. 2001a, 2001b). For German, we did not already have a sentence realization module. We therefore embarked on the undertaking described in this technical report, namely to produce an empirically-based sentence realization module by employing machine learning techniques as much as possible. As a first step towards a generation module that is usable in machine translation contexts, we created a module that generates German output strings from German input strings by roundtrip of analysis and subsequent generation as a proof-of-concept. The main focus of this report is on the German-to-German generation approach, although we briefly discuss first results of the application of the Amalgam system in machine translation in section  REF _Ref8029913 \r \h 10. The advantage of approaching the task from the German-to-German generation perspective is that evaluation of the system is straightforward. The input string goes through syntactic and semantic analysis, a logical form representation is produced, and an output string is generated from that logical form representation by Amalgam. If the output string is identical to the input string, Amalgam has performed flawlessly. The farther the output string is from the input string, the worse Amalgam has performed. This is, of course, an over-simplified view, given that there is often more than one good German sentence that would represent a logical form faithfully and fluently (see, for example, the discussion of relatively free constituent order in German in section  REF _Ref8030558 \r \h 3.4). Despite this caveat, the German-to-German approach has provided a good starting point for the development of the first prototype of Amalgam. Amalgam has been described in published papers (Corston-Oliver et al. 2002, Gamon et al. 2002a, Gamon et al. 2002b). Our goal in this technical report is to provide a complete description of the architecture and implementation of Amalgam, beyond the level of detail that is customary in conference proceedings or journal papers. We hope that this technical report will provide answers to some of the questions that inevitably arise when reading published descriptions of natural language processing systems, including the following questions: Exactly which features are used? How are the features extracted? How much work is performed by the knowledge-engineered module mentioned in passing? Prior work in sentence realization Reiter (1994) surveys the major natural language generation systems of the late 1980s through the mid-1990s: FUF (Elhadad 1992), IDAS (Reiter at al. 1992), JOYCE (Rambow and Korelsky 1992), PENMAN (Penman 1989) and SPOKESMAN (Meteer 1989). Each of these systems has a different theoretical underpinning: unification grammar in the case of FUF (Kay 1979), a generalized reasoning system (Reiter and Mellish 1992) in the case of IDAS, Meaning-Text theory (Mel uk 1988) in the case of JOYCE, Hallidayan Systemic Functional Linguistics (Halliday 1985) and Rhetorical Structure Theory (Mann and Thompson 1988) in the case of PENMAN, Tree-Adjoining Grammar (Joshi 1987) in the case of SPOKESMAN. Despite their diverse theoretical underpinnings, Reiter draws attention to the fact that a consensus appeared to have emerged concerning the appropriate architecture for a natural language generation system. All systems had a module that performed content determination, mapping input specifications of content onto a semantic form, followed by a module that performed sentence planning. Output was performed by a surface generation module (what we refer to below as a sentence realization module) that made use of a morphology module and a module that performed formatting of the output text. All the systems that Reiter surveyed generate English text only. Reiter draws speculative analogies between the consensus architecture and the evidence for modularity of language in the human brain based on language impairment of individuals with various types of brain injury, and suggests that the engineering trade-offs made during system implementation might mirror evolutionary forces at work in the development of human language. The dominant paradigm for natural language generation systems up until the mid-1990s was that of knowledge engineering. Computational linguists would explicitly code strategies for stages ranging from planning texts and aggregating content into single sentences to choosing appropriate forms of referring expressions performing morphological inflection and formatting output. This research path has yielded several mature broad-coverage systems and is still being actively pursued today; see, for example, the sentence realization modules described in Aikawa et al. (2001). Since the mid 1990s there has been increasing interest in the application of statistical and machine learning techniques to the various stages of natural language generation. This research has ranged from learning plans for high-level planning of texts and dialogues (Zukerman et al. 1998, Duboue and McKeown 2001) or ensuring that the macro properties of generated texts such as the distribution of sentence lengths and lexical variety mirror the properties of naturally occurring texts (Oberlander and Brew, 2000) to sentence planning (Walker et al., 2001), lexical selection (Bangalore and Rambow 2000b), selection of the appropriate form for a referring expression (Poesio et al 1999), determining grammatical relations (Corston-Oliver 2000) and selecting the appropriate word order (Langkilde and Knight 1998a, Langkilde and Knight 1998b, Bangalore and Rambow 2000a). It is often the case in the natural language generation literature that descriptions of the higher level aspects of natural language generation gloss over issues associated with sentence realization. Walker et al. (2001), for example, characterize realization in this way: During realization, the abstract linguistic resources chosen during sentence planning are transformed into a surface linguistic utterance by adding function words (such as auxiliaries and determiners), inflecting words, and determining word order. This phase is not a planning phase in that it only executes decisions made previously. (Walker et al. 2001 ) In typical implementations, however, once the planning stages, sensu stricto, have finished there remain myriad encoding decisions to be made and selections among alternative formulations to be performed. Increasingly, machine-learned techniques are being brought to bear on these tasks. Two recent systems, the Nitrogen system (Langkilde and Knight 1998a, Langkilde and Knight 1998b) and the FERGUS system (Bangalore and Rambow 2000a) are sufficiently similar to Amalgam to warrant extended discussion. The Nitrogen system (Langkilde and Knight 1998a, Langkilde and Knight 1998b) uses word bigrams instead of deep symbolic knowledge to decide among alternative output sentences. The input to Nitrogen can range from rather abstract semantic representations to more fully-specified syntactic representations. Inputs are given in the Abstract Meaning Representation, based on the Penman Sentence Plan Language (Penman 1989). Two sets of knowledge-engineered rules operate on the input specification to produce candidate output sentences. One set of rules performs one-to-many mappings from underspecified semantic representations to possible syntactic formulations, fleshing out information such as definiteness and number that might be missing in practical generation contexts such as Japanese to English machine translation (Knight et al 1995). The second set of rules, which includes sensitivity to the target domain, transforms the representations produced by the first module to yield still more candidate sentences. The candidate sentences are compactly represented as a word lattice. Word bigrams are used to score and find the optimal traversal of the lattice, yielding the best-ranked output sentence. Morphological inflection is performed by simple table lookup, apparently during the production of candidate sentences. Langkilde and Knight (1998a) present worked examples that illustrate the importance of the bigram filtering. One input semantic form includes five lexical nodes in such relationships as Agent, Destination, and Patient. The word lattice that results contains more than eleven million possible paths, with the top-ranked candidate Visitors who came in Japan admire Mount Fuji. Another worked example, for which the semantic representation is not given, appears to involve two content words that are transformed into a word lattice containing more than 155,000 paths to yield the top-ranked candidate I cannot betray their trust. Clearly, there is an important interaction between the knowledge-engineered components that propose candidates and the bigram filtering. If the knowledge-engineered components are too conservative, an optimal rendering will not be proposed, forcing the bigram filtering component to choose among sub-optimal candidates. On the other hand, if the knowledge-engineered component proposes too many candidates, the search through the lattice may become so time-consuming as to be impractical. Unfortunately, Langkilde and Knight do not give more details about the knowledge-engineered components. One wonders how many rules there are, how many rules must be added for a new domain, and what level of expertise is required to write a rule. The use of bigrams is problematic, as Langkilde and Knight acknowledge. Bigrams are unable to capture dependencies among non-contiguous words, a fact that is perhaps mitigated by the observation that in practice, for English at least, syntactic dependencies most often obtain between adjacent elements (Stolcke 1997). Increasing the number of terms to trigrams or higher-order n-grams raises the familiar specter of paucity of data. Furthermore, as Langkilde and Knight observe, many linguistic relationships are binary in nature, and therefore not efficiently represented using trigrams. To overcome some of the deficiencies of the bigram language model applied to a word lattice, Langkilde (nd) proposes using a parse forest to represent the output candidates. The evaluation metric that she intends to develop would combine information from the syntactic representations and the language model. It is unclear how feasible it would be to generate candidate sentences and then filter them when generating German. For English, Langkilde and Knight use table lookup to add morphological variants of content words to the mix. Since English inflectional morphology is relatively simple, this does not adversely explode the search space. When we consider the richer inflectional morphology of German, however, this simple table lookup does not appear practical. Whereas English has a single definite article, the, German has six inflected forms (der, die, das, etc). Similarly, English adjectives can be inflected only for degree (e.g., big, bigger, biggest), whereas German adjectives additionally distinguish three lexical genders, four cases, and singular vs. plural. If the search space becomes large using table-based lookup to propose additional nodes in a word lattice for English, it would become intractable for a language such as German with richer inflectional morphology. Recent research has demonstrated the usefulness of syntactic information in overcoming the inadequacies of bigrams. Ratnaparkhi (2000) demonstrates dramatic improvements in selecting appropriate output templates for the air travel domain when conditioning on syntactic dependencies versus conditioning on trigrams. Further validation of the usefulness of syntax during sentence realization can be seen in the Fergus system (Bangalore and Rambow 2000a). Bangalore and Rambow augment the work of Langkilde and Knight by adding a tree-based stochastic model and a traditional tree-based syntactic grammar. Bangalore and Rambow take as input a dependency tree. A stochastic tree model chooses TAG trees for the nodes in the dependency tree. The resulting TAG analysis is then unraveled to produce a lattice of compatible linearizations. Selection among competing linearizations is performed by a Linear Precedence Chooser which selects the most likely linearization given a suitable language model. To date there have been no published descriptions of the application of machine learning to the problems of morphological realization or formatting for natural language generation, although presumably inflectional morphology that had been learnt automatically (e.g., Goldsmith 2001) could subsequently be applied during generation. Properties of German The German language exhibits a number of properties that are very different from English, despite the fact that the two languages are relatively closely related. These properties pose challenges to a sentence realization component which go beyond what an English sentence realizer would have to account for. For us, this poses the interesting task of making the overall design of Amalgam flexible enough to deal with these phenomena, and as a result, flexible enough to be applicable to more languages. It also protects us from the myopia of NLP solutions predicated on properties of English, such as the paucity of inflectional morphology and the relative rigidity of constituent order. In this section, we present a brief overview of the important characteristics of German. It should be understood that this is by no means a complete list of the properties in which German differs from English. We focus on a handful of properties crucial in sentence realization. We contrast these properties with English, to emphasize the typological differences between the two languages, particularly for the benefit of English speakers who are not familiar with German. The Position of the Verb in German One of the most striking properties of German, painfully familiar to anyone who has learned German as a foreign language, is the distribution of verbs in main and subordinate clauses. In fact, most descriptive accounts of German syntax are based on a topology of the German sentence that treats the position of the verb as the fixed frame around which other syntactic constituents are organized in a relatively free order (cf. Eisenberg 1999, Engel 1996). The general frame of the German sentence is shown in  REF _Ref8723579 \h Figure 1.  EMBED Visio.Drawing.4  Figure  SEQ Figure \* ARABIC 1: The topological model of the German sentence The important facts to note about this topological model are: The Prefield contains at most one constituent The Left Bracket contains: the finite verb OR a subordinating conjunction OR a relative pronoun/relative expression The Middle Field contains any number of constituents The Right Bracket contains all the verbal material that is not present in the Left Bracket. If the finite verb is in the Left Bracket, then the Right Bracket contains the non-finite verbs. If the Left Bracket is occupied by a subordinating conjunction or relative expression, the Right Bracket contains all the verbs. The Postfield contains: clausal complements subordinate clauses extraposed material (e.g., relative clauses extraposed from the Middle Field) other constituents The position of the verb in German is rigidly fixed. Errors in the positioning of the verb will result in gibberish, while most permutations within Prefield, Middle Field and Postfield will at worst result in less fluent output. Depending on the position of the finite verb, German sentences and verb phrases are often classified as being verb-initial, verb-second or verb-final. In verb-initial clauses, the finite verb is in initial position (e.g., in the imperative example in  REF _Ref530806740 \h Figure 2). Verb-second sentences contain material in the Prefield, and a finite verb in the Left Bracket. Verb-final sentences (such as the complement clause in  REF _Ref530806740 \h Figure 2) contain no verbal element in the Left Bracket (usually because the Left Bracket is occupied by a subordinating conjunction or a relative pronoun).  REF _Ref530806740 Figure 2 illustrates the above generalizations with some examples. German text is in italics, glosses are given below each word, and free translations are given in quotes. PrefieldLeft BracketMiddle FieldRight BracketPostfieldMain clauses (declarative)Hans Hanssieht seesdas Auto the carHans sees the carHans Hanshat hasdas Auto the cargesehen seenHans has seen the carHans Hansgibt givesdas Buch the bookab PrefixHans returns the bookHans Hanswird willdas Auto the cargesehen haben seen haveHans will have seen the carHans Hanshat hasdas Auto the cargesehen seendas er kaufen mchte that he buy wantsHans has seen the car that he wants to buyMain clauses (interrogative)Was Whatsieht seesHans HansWhat does Hans see?sieht seesHans das Auto Hans the carDoes Hans see the car?Main clauses (imperative)sieh seedas Auto the carSee the car!Complement clausesdass thatHans das Auto Hans the cargesehen hat seen hasthat Hans has seen the carRelative clausesdas whichHans Hansgesehen hat seen hasthat Hans has seenFigure  SEQ Figure \* ARABIC 2: Examples of the topological model applied to German sentences Separable Prefixes A large percentage of German verbs fall in the class of separable prefix verbs (in the Nlpwin lexicon, roughly 8,000 of a total of 20,000 verbs fall in this category). The peculiarity of these verbs is that they form a semantic unit, but are separated syntactically into two parts, one of which is a finite verb stem, the other is a prefix that occupies the position of a non-finite verb in the topological model of the sentence. Consider the example abgeben which is the German verb meaning return. This verb consists of two parts, a prefix ab and a verb stem geben. The semantics is not compositional, although there certainly is at least some overlap between the meaning of the stem geben to give and the separable prefix verb abgeben. In verb-second clauses such as the declarative main clause in  REF _Ref530806740 \h Figure 2, the stem and the prefix separate, with the stem occupying the Left Bracket, and the prefix occupying the Right Bracket: Hans gibtStem das Buch abPrefix Hans returns the book The correct positioning of prefix and stem is an integral part of sentence realization in German. Any simple-minded mapping of word-to-word in machine translation, for example, will fail miserably if the target language is German unless some mapping from one verb in English to both a prefix and a stem in German is possible, and their correct positioning in the topological model is ensured. Morphological Case German has a rich system of inflectional morphology. Particularly important for sentence realization as well as parsing in German is case marking on noun phrases. There are four cases in German: nominative, accusative, dative, and genitive. Depending on a number of factors such as the morphological class of the lexical items, number, gender, and the choice of determiner, case can be morphologically realized on various elements of the noun phrase: the noun itself, and (if present) determiners and adjectives. Case is often an important clue in determining the semantic role of a noun phrase in the clause. If an active clause contains a nominative and an accusative noun phrase, the nominative phrase can safely be assumed to be the subject, and the accusative phrase to be the object, independently of their linear order in the sentence string. Constituent Order The ordering of words and constituents varies across languages, and so does the rigidity with which the canonical order must be obeyed. We will restrict ourselves to the discussion of free constituent order, since neither English nor German can be reasonably claimed to exhibit any free word order in the real sense; i.e., neither English nor German show examples where individual words can be ordered freely, outside of the immediate constituent that they belong to. English has a relatively rigid constituent order although a number of preposing and extraposing operations can alter that order, so that any simplistic claim about fixed constituent order in English is problematic. German, on the other hand, allows many major constituents to be rather freely distributed amongst Prefield and Middle Field, and to a somewhat lesser extent in the Postfield. At the same time, the position of the verb is fixed to the two bracket positions as described in section  REF _Ref530816124 \r \h 3.1. Below are some examples to illustrate this point (English glosses at the bottom of the example): [Unter diesen Umstnden] hat [die Firma] [weitere Lieferungen] bestellt, [ohne abzuwarten]. [Die Firma] hat [unter diesen Umstnden] [ohne abzuwarten] [weitere Lieferungen] bestellt [Weitere Lieferungen] hat [die Firma] [unter diesen Umstnden] bestellt, [ohne abzuwarten] [Ohne abzuwarten] hat [unter diesen Umstnden] [die Firma] [weitere Lieferungen] bestellt Gloss: Under these circumstances the company has ordered further shipments without waiting. [ohne abzuwarten] = without waiting [unter diesen Umstnden] = under these circumstances [die Firma] = the company [weitere Lieferungen] = additional shipments [... hat ... bestellt] = has ordered All of the above permutations and many more logically possible ones yield grammatical German sentences. At the level of predicate argument structure, the meaning of all the permutations is identical. At finer-grained levels of semantic/ pragmatic description, such as theme/rheme, topic/focus, background/foreground information, there clearly are differences, however. Since none of these finer grained distinctions are currently computable or representable in the Nlpwin framework, we will ignore them for the remainder of this report. Extraposition of Clauses In both German and English, it is possible to extrapose clausal material to the right periphery of the sentence, as the following examples illustrate. Relative clauses: English: The man entered the room who usually causes trouble right away. German: Der Mann hat den Raum betreten, der blicherweise immer rger macht. Infinitival clauses: English: The possibility was considered to leave the country. German: Man hat die Mglichkeit erwogen, das Land zu verlassen. Complement Clauses: English: A rumor has been circulating that he is ill. German: Ein Gercht ging um, dass er krank ist.  REF _Ref530822123 \h Figure 3 shows that English and German differ in the frequency of this phenomenon. The results shown are based on automatic data profiling with Nlpwin, where the output of the parser has been postprocessed to indicate relative clauses (RELCL), infinitival clauses (INFCL), or complement clauses (COMPCL) that have been moved from their original position. The analysis is based on 100,000 aligned English-German sentence pairs from Microsoft technical manuals and a different German corpus of 62k sentences consisting of a mixture of news, grammar book examples, user input and other sources. Nearly one third of German relative clauses are extraposed in technical writing, while only 0.18% of English relative clauses are extraposed in the corresponding sentence set. For infinitival clauses and complement clauses the numbers are more comparable between English and German. The high number of extraposed relative clauses in German accords with the number reported in Uszkoreit et al. (1998), who observe 24% of relative clauses being extraposed in a hand-annotated German news corpus. German technical corpus (100K sentences)English technical corpus (100K sentences)German balanced corpus (62K sentences)RELCL32.96%0.18%24.06%INFCL5.56%0.82%4.26%COMPCL2.26%4.61%2.81%Figure  SEQ Figure \* ARABIC 3: Percentage of extraposed clauses Since extraposition is so rare in English, an English sentence realization module could safely ignore extraposition and still result in very fluent output. A complete German sentence realization module, however, will need to model extraposition. The modeling of extraposition is discussed further in Gamon et al. (2002b). The Nlpwin system The syntactic analysis: Sketch and Portrait Nlpwin produces two levels of syntactic output: an initial constituent analysis in which attachment ambiguities are represented in a packed tree (the Sketch), followed by a constituent analysis in which attachment ambiguities have been resolved (the Portrait). Syntactic trees in Nlpwin are flattened representations of a syntactic analysis in terms if binary augmented phrases structure rules (see Jensen et al. 1993 for more details on the formalisms). Each syntactic node in Nlpwin has a head, and it may have pre-modifiers and/or post-modifiers.  REF _Ref531763254 \h Figure 4 illustrates a syntactic structure for a German sentence. Nodes in this structure correspond to attribute-value data structures of considerable complexity.  REF _Ref531763258 \h Figure 5 shows the record for the node NP2 in  REF _Ref531763254 \h Figure 4.  Figure  SEQ Figure \* ARABIC 4: Example of a syntactic structure produced by Nlpwin  Figure  SEQ Figure \* ARABIC 5: Example of a syntactic record in Nlpwin The semantic representation: Logical Form Subsequent processing in Nlpwin computes a semantic representation, the Logical Form (LF). The Logical Form is a graph data structure that represents the core predicate argument structure and basic semantic relations. Semantic relations are encoded as labeled arcs between semantic nodes. Semantic nodes are lexical in the sense that they are derived from syntactic nodes and are labeled with the citation form of the head of the syntactic node from which they were derived. There are no abstract semantic nodes in the Nlpwin LF. An example of a logical form graph for the English sentence You have to click on the tab in order to print the document on the printer is given in  REF _Ref531749592 \h Figure 6.  Figure  SEQ Figure \* ARABIC 6: An example of a Logical Form graph Logical form nodes can carry many features, some of which are related to lexical properties of the associated lemma (e.g., Conc on tab1 in  REF _Ref531749592 \h Figure 6), others are semantic features based on the particular analysis (e.g., Pres for present tense on the node click1). Only content words are represented in logical form, where a content word is understood to be a word that cannot be represented as a small set of features or a label on a semantic arc. The total number of different semantic relations in the Nlpwin system is relatively small; the current system has about 40 semantic relations. Some examples of semantic relations are given below: Basic predicate argument structure relations: Tsub, Tobj, Tind (for semantic subjects, objects, and indirect objects) Other semantic relations: Means, Time, Duration, Locn (location), Cause, Mannr (manner), Purp (purpose), Result, Measure, Classifier, Equiv (equivalence), Possr (possessor), LTopic (topic), Props (propositions), Mod (unspecified modifiers) Semantic nodes in logical form contain pointers to the syntactic nodes they were derived from. This is an important bookkeeping device, enabling us to train models on the correspondences between syntactic and semantic nodes in order to learn the conditions for the operations necessary to transform one into the other. The procedural flow The Major Stages of the Amalgam Pipeline The Amalgam pipeline consists of eight stages, which perform linguistically distinct sets of operations, as shown in  REF _Ref11047897 \h Figure 7. Pre-Processing (preparing the LF graph for further processing) Degraphing Addition of lexical information through dictionary lookup Simplification of German compounds Flesh-out (adding syntactic information to the LF graph) Addition of syntactic labels Insertion of function words such as determiners, auxiliaries, semantically empty prepositions, relative pronouns, reflexive pronouns, etc. Assignment of spellout probabilities for NPs in subject or object position Assignment of morphological case features Assignment of verb-position features Conversion to basic tree: Reading off a syntactic tree structure from the degraphed LF Splitting of separable prefixes from their stems Introduction of syntactic representation of coordination Reversing of certain syntactic dominance relations Global movement: Raising, Wh-movement and movement of relative pronouns Extraposition Setting of underspecified verbal inflectional features (agreement, participial features, etc.) Intra-constituent ordering (establishment of linear order) Surface cleanup: Surface realization of determiners, relative pronouns and reflexive pronouns Deletion of duplicated material in coordination (syntactic aggregation) Punctuation Inflectional generationFigure  SEQ Figure \* ARABIC 7: Overview of the Amalgam pipeline In general, the order of the stages reflects dependencies in the pipeline. For example, punctuation depends on an established order of constituents, and so does deletion of duplicated material in coordination. Movement, as we treat it in Amalgam, is an operation that moves a constituent out of its parent constituent and attaches it higher in a syntax tree without establishing linear order so the ordering of constituents has to follow hierarchical movement. Assignment of case features, verb-position features, and syntactic labels is best performed before the basic tree is established, in order to provide a complete syntactic representation and helpful information for following stages. Syntactic labels in particular are an important source of information for models downstream in the pipeline. A few steps could be performed equally well at different points in the Amalgam pipeline. Function words, for example, are inserted during Flesh-out in Stage 2. We decided that the insertion of function words was conceptually similar to the other operations performed during Flesh-out that augment the Logical Form with syntactic information. We believe that the order of the major stages is highly language-independent. The order of operations within each stage is not significant. In section  REF _Ref530829785 \n \h 5.1.1- REF _Ref530829811 \n \h 5.1.8 we will illustrate the workings of these eight stages through screenshots of the corresponding structures/strings on the basis of the example sentence Hans isst die Kartoffeln auf, die er gestern geernet hat Hans eats up the potatoes which he has harvested yesterday. Not all of the individual operations are performed in this one sentence, but it serves as a general illustration of the kinds of operations in the Amalgam pipeline. In sections  REF _Ref530882302 \n \h 6 and  REF _Ref530882316 \n \h 7 we discuss each of the procedural and machine-learned operations in more detail. For this example, the input LF graph to the generation component is given in  REF _Ref531749173 \h Figure 8. To simplify the display of the graph, the display algorithm attempts to minimize crossing lines. The node Kartoffel1 is displayed twice in the graph but is in fact the same node, i.e. Kartoffel1 is dependent on aufessen1 and ernten1. Similarly, Hans1 is dependent on aufessen1 in the Tsub relation and is a possible intrasentential coreferent of er1.  Figure  SEQ Figure \* ARABIC 8: LF structure for the sentence Hans isst die Kartoffeln auf, die er gestern geernet hat Pre-Processing In the Pre-Processing stage, the LF graph is degraphed; i.e., a structure is created in which each node has at most one parent node. This operation creates a tree structure that facilitates conversion into a syntactic tree in the subsequent stages. Nodes that needed to be duplicated in order to create a tree from the graph bear coindices which link them to their counterparts. In addition, a lexical lookup in the German Nlpwin dictionary is performed on the lexical items present in the graph, and the dictionary information is stored in an attribute on the records in the graph (not displayed in  REF _Ref530883055 \h Figure 9). Finally, compound nouns that have been analyzed into components in LF are reconverted into an un-analyzed compound string for training and generation purposes. The output of the pre-processing stage is illustrated in  REF _Ref530883055 \h Figure 9. Note that the nodes Kartoffel1 and Hans1 which have previously been in two dependency relations have each been duplicated. Kartoffel1, for example, is now only dependent on aufessen1. Kartoffel1 has been cloned to produce Kartoffel2, which is dependent on ernten1. The coindices are shown in  REF _Ref530883055 \h Figure 9. A positive coindex denotes the node from which another node was cloned. A negative coindex indicates that the node was originally cloned from the node whose coindex is the corresponding positive integer, e.g., Karotoffel1 (coindex of 2) is the node from which Kartoffel2 (coindex of -2) was derived.  Figure  SEQ Figure \* ARABIC 9: The output of the pre-processing stage Flesh-out During the Flesh-out stage, information is added to the degraphed LF. Typically, these are details about syntactic realization that have been normalized at the more abstract LF level of representation. First, syntactic labels are assigned to the nodes in the degraphed LF, based on a decision tree classifier (the new syntactic labels are present in an attribute on the nodes, but are not displayed in  REF _Ref530884156 \h Figure 10 below). Function words that carry no semantic information are not present at LF. These function words are inserted next and include: An abstract determiner: Defdet for definite determiners, Indefdet for indefinite determiners, Whdet for Wh determiners, and Proxldet and Distldet for demonstrative determiners. The surface form of these determiners is determined later, during the surface cleanup stage. Auxiliaries. Prepositions which have a purely syntactic function. In German, this includes the prepositions von and durch used in the passive construction. The infinitival marker zu. Negation nicht, which is marked in the LF as the feature [+Neg]. Subordinating conjunctions dass and ob. Expletive subjects, i.e. the semantically empty grammatical subject es. An abstract relative pronoun Relpro which receives its surface realization during stage 6, surface cleanup. Reflexive pronouns. The Wh adverbial wie how. The function words given in 1-7 are inserted based on a decision tree classifier for each of the insertion tasks. The function words given in 8-10 are inserted by simple functions. In addition to the insertion operations, a function contracts LF nodes of prepositional proforms such as dadurch, damit ("through that", "with that") etc. to their surface string. Logical subjects and objects are assigned a probability for spellout by a decision tree classifier, i.e., a probability of their being realized in the surface string. Logical subjects of infinitival clauses, for example, should not be overtly represented in the string. Finally, case features and verb position features are assigned by decision tree classifiers.  REF _Ref530884156 \h Figure 10 shows the result of the flesh-out operations on our sample LF.  Figure  SEQ Figure \* ARABIC 10: The degraphed LF after Flesh-out Conversion to basic tree During conversion to a basic tree, the first operation is the actual removal of logical subjects and objects which have a low probability of overt realization (as assigned by a decision tree classifier during Flesh-out). The degraphed LF at this point is transformed into a syntactic tree structure. The syntactic labels on nodes in the degraphed LF that were assigned during Flesh-out are copied over to the corresponding non-terminal nodes in the basic tree. Separable prefixes are split from their stem, based on verb-position features assigned in the previous stage, and based on lexical information about the boundary between the prefix and the stem obtained from the dictionary during Pre-processing. In the next two steps, the representation of coordination is mapped from the way it is handled in LF (see section  REF _Ref530885112 \r \h 4 and section  REF _Ref11560948 \r \h 6.2.9) to a more surface oriented structure in which the coordinating conjunction is the syntactic head. The last step in the conversion to basic tree is an operation based on a decision tree classifier which reverses syntactic dominance relations in those contexts where syntactic and semantic dominance relations are at odds, particularly in cases involving quantification, e.g., viele der Leute many of the people where viele many is the syntactic head, but Leute is the semantic one.  REF _Ref530885349 \h Figure 11 shows the basic tree structure for the example sentence. In parentheses on the far-right, the LF relations of the nodes to their semantic parent are displayed, relations starting with a tilde denote pseudo-relations, i.e., inserted material that had no original place in the LF and hence no original LF relation. The presence of the LF relations is a reminder that the new syntactic nodes in the basic tree bear references to the LF nodes from which they were constructed. The LF features continue to be accessible to downstream modules.  Figure  SEQ Figure \* ARABIC 11: Basic tree structure for the example sentence Global movement During Global Movement, non-local movement operations are performed. Non-local here means movement beyond the limits of the immediate parent. All local movement in Amalgam is treated as an ordering phenomenon within one constituent, not as genuine movement. Raising, Wh-movement, and the movement of relative pronouns/relative expressions are handled by three simple functions. While this seems strangely at odds with the attention that these operations have received in linguistic research, it is important to note that the more involved examples of multiple Wh-movement, long distance Wh-movement, parasitic gaps, etc., which are important phenomena from a theoretical point of view, are extremely rare in real-life texts. Given the rarity of these phenomena in our training set, we decided to deal with these phenomena with rules. In principle, a machine-learned approach could be applied, given sufficient training examples. The next movement step, extraposition of relative, infinitival and complement clauses, is based on a decision tree classifier which decides for each instance of such a clause whether it should move up one step and attach to the parent of its parent. Once reattached there, the next hop is evaluated, until a position is found where the probability of further movement is less than the probability of no further movement (and hence less than 0.5). A trace is left behind in the original position, with a pointer to the extraposed clause (and vice versa). The final steps in the global movement stage are two functions which assign morphological features for verbs based on the information present in the tree nodes. Tense information is copied to the finite verb which might be an auxiliary inserted during flesh-out. Participial information is copied onto the non-finite verb, and agreement of the finite verb with the subject is established. The surface subject is identified as the first nominative NP in the domain of the finite verb (case features having been assigned during flesh-out by a decision tree classifier). In our example, as shown in  REF _Ref530886540 \h Figure 12, extraposition of the relative clause RELCL3 has taken place.  Figure  SEQ Figure \* ARABIC 12: The tree after Global Movement Intra constituent ordering During this stage, a generative language model of syntax tree structure is applied in a beam search to establish the linear order of the nodes within each constituent, and consequently the tree. The model consists of n-gram probabilities (currently, n=2) on the order of the labels of the nodes and the semantic relations of the nodes (from the LF), conditioned on constituent features, such as parent and head nodetypes (for more details see section  REF _Ref530886833 \n \h 7.2 below). Currently, we apply a fix-up function after the application of the model to correct the positions of the verbs, based on the verb position features assigned by a decision tree classifier in the Flesh-Out stage. We plan to eliminate this function as work on the order model progresses. As currently structured, verb position cannot be reliably established by the order model, due to the limitations of an n-gram window unconditioned on verbal features. For any given n, an n-gram is inadequate for capturing the generalizations about verb position where the finite verb and non-finite verbs in a verb-second structure can be separated by a theoretically unlimited number of constituents. One possible extension to the model that we envision is the addition of a separate model of the verb position. At the end of the ordering stage we apply a function that assembles compounds from LFs containing nouns with non-possessive noun modifiers. This function is currently only used in machine translation contexts (recall from section  REF _Ref530829785 \r \h 5.1.1 that we simplify compounds when processing German-to-German generation). In our example, ordering results in the tree in  REF _Ref10887249 \h Figure 13.  Figure  SEQ Figure \* ARABIC 13: The ordered tree Surface cleanup It is obvious from  REF _Ref10887249 \h Figure 13 that some work still needs to be done to that structure in order to arrive at a correct surface string. Both the abstract relative pronoun Relpro and the abstract determiner Defdet need to be converted to their surface realization. These tasks are achieved during surface cleanup by two decision tree classifiers which decide on the most probable surface form. Note that in German this is not a trivial task: during training, the model has picked up on no less than 55 different determiner forms from the training corpus, and 23 different forms of relative pronouns. Reflexive pronouns, which also received an abstract form during insertion in flesh-out, are converted into their surface form by a simple function. The result of these operations for our example sentence is shown in  REF _Ref530888667 \h Figure 14.  Figure  SEQ Figure \* ARABIC 14: The tree after surface cleanup Surface cleanup contains an additional step not illustrated in the example sentence: the reduction of duplication in coordinated constituents, also called syntactic aggregation in the generation literature. Consider a sentence like Hans hat die Kartoffeln gekocht und gegessen Hans has cooked and eaten the potatoes. The LF for this sentence correctly establishes semantic relations between each of the verbs kochen and essen and the subject Hans and object die Kartoffeln. Mapped to a tree through Amalgam, the surface string will faithfully encode all the relations that were present in the input LF, resulting in duplication: Hans hat die Kartoffeln gekocht und Hans hat die Kartoffeln gegessen Hans has cooked the potatoes and Hans has eaten the potatoes. Although this is a perfectly grammatical German sentence, it is not the desired fluent output we would wish to produce. Surface cleanup contains two operations dealing with syntactic aggregation. The first operation is based on a decision tree classifier which establishes a probability of being overtly realized for each of the duplicated nodes in a coordination structure. Each of the duplicated nodes with p(overtly realized) > 0.5 will be retained, while the duplicated nodes with lower probability are eliminated. In case there is not a single duplicated node that reaches the probability threshold, the node with the highest probability of being realized is retained as a safeguard against truncation. The second operation is a function that eliminates duplicated function words such as prepositions and auxiliaries. Punctuation After the creation of an ordered and fully spelled-out tree, punctuation needs to be inserted to ensure fluent and readable output. Punctuation rules are notoriously difficult in German, and although some simplification has been achieved in the spelling reform, there are still 26 different rules for the correct positioning of the comma alone. Since punctuation conventions are typically in the form insert punctuation X after Y or insert punctuation X before Y, we decided to build two different decision tree classifiers for preceding and for following punctuation. We only train and apply these models for sentence internal punctuation, since sentence final punctuation can be inserted with a simple function. At each terminal node in the tree, the left edge of that terminal node and the right edge of the preceding node are passed into the classifier for preceding and following punctuation, respectively. The verdicts from both classifiers are collected and if there is any strong prediction (>0.5) for the insertion of punctuation, the strongest such prediction wins and the predicted punctuation mark is inserted. In our example, one comma is inserted before the extraposed relative clause, as shown in  REF _Ref530892005 \h Figure 15.  Figure  SEQ Figure \* ARABIC 15: Tree with inserted punctuation Inflectional generation The final stage in the Amalgam pipeline is inflectional generation. The records in the tree structure at this stage in the pipeline contain all necessary information to be passed into a rule-based inflectional morphology component for German. This component has been developed for use in the German grammar checker in the Microsoft Word word processor. Features passed into the inflectional generation component include case, gender, number, person, etc. To give an example, the record of the node STEM1 in the tree in  REF _Ref530892005 \h Figure 15 is shown in  REF _Ref530901010 \h Figure 16. Based on the features Pers2, Sing, Pres, and Indicat, the verb form isst can be generated from the Lemma essen.  Figure  SEQ Figure \* ARABIC 16: Attribute-value data structure for the finite stem in the example sentence  REF _Ref530901197 \h Figure 17 contains the final result of the generation process on the example sentence, including the inflected forms of the verbs essen, ernten, and haben and of the noun Kartoffel.  Figure  SEQ Figure \* ARABIC 17: The final inflected tree A string is read off this final tree structure, and in our case, the output string corresponds exactly to the input string: Hans isst die Kartoffeln auf, die er gestern geerntet hat. A Detailed Flowchart of the Amalgam Pipeline  EMBED Visio.Drawing.6   EMBED Visio.Drawing.6  EMBED Visio.Drawing.6  Figure  SEQ Figure \* ARABIC 18: A detailed flowchart of the Amalgam pipeline The rule-based operations in Amalgam Degraphing We begin with a logical form graph as input to the generation process. The first step in producing a linear sequence of words is to disentangle the logical form graph to produce a degraphed logical form. Each node in the input logical form contains a list of pointers to parent nodes, stored in the attribute Parents. A parallel list of atoms, ParentAttrs, stores the corresponding labels on the arcs to those parent nodes. These attributes are updated during the degraphing. When the degraphing is complete, every node but the root has exactly one parent. The root has no parent, by definition. Nodes that are replicated during degraphing are assigned a numerical index, stored in the CoIndex attribute. One node is assigned a positive integer, while the duplicates are assigned a negative integer. During the degraphing operation, certain logical form attributes are ignored (i.e., their values are not cloned). These fall into three classes: System-internal bookkeeping attributes of no linguistic interest, e.g., CopyOf, a pointer to the record that the current record was copied from during construction of the logical form. The other bookkeeping attributes are Originl, Clones, CopyLFCopiedTo, Rules, Constits, CopyOf, BoxCodeChecks, and LexNode. Attributes that used to encode the restructured logical form. These are ParentAttrs, Parents, and CoIndex. Attributes used only for advanced semantic processing, but not yet considered reliable or useful for generation. These include attributes indicating intra-sententential coreference (Refs, RefOf) and the attributes used for MindNet, a semantic knowledge base (Richardson et al. 1998), namely WeightedPaths, MatchPaths, Topicl, Simples, AmbRecs, ExpandSCs, AmbGCs, Counts, Masses, Coordnode, HypSynLems, Emph, and Nominf.  REF _Ref531761766 \h Figure 19 shows the logical form for the sentence Alle Artikel und Publikationen jeder ausgewhlten Datenbank werden bertragen und aktualisiert All articles and publications of each selected database are transmitted and updated. This logical form contains a fair number of nodes with multiple parents, which require degraphing in Amalgam.  REF _Ref531762011 \h Figure 20 illustrates the same logical form after degraphing.  Figure  SEQ Figure \* ARABIC 19: German logical form before degraphing  Figure  SEQ Figure \* ARABIC 20: German logical form after degraphing Miscellaneous rule-based operations Creation of lexnodes The function create_lexnodes performs a lookup of the lemma of a node in the logical form. It then stores the information retrieved from the dictionary in an attribute on that node. The purpose is to make lexical information (such as subcategorization information) available to the subsequent processing stages. For hyphenated words or compounds, the function performs analysis of the word and returns lexical information of the head of the compound or hyphenated item. Simplification of compounds German compounding (especially nominal compounding) is very productive. The German analysis system contains a compounding analysis module, which tries to identify the parts of a compound such as Eingangsbereich (entry area) by using lexical information, word frequency information and syllable structure restrictions. The correct analysis in this example is Eingang + s + Bereich, where s is what is called a linking morpheme. At the level of logical form, the head of the compound forms a node in the LF graph (in this case Bereich), and the other meaning-bearing parts of the compound are linked to that node through the Mods attribute. For the purposes of Amalgam training, we ignore the internal structure of compounds, which is part of morphological generation, not syntactic generation. For training purposes, we simplify the logical form representation, effectively undoing the compound analysis. The result of that process on a node of a compound word is a single node with the complete compound string. Contracting PPs German has an array of PP proforms such as damit, dafr, etc. These forms contain a preposition (mit, fr in the example) and a pronominal element da. At LF, we currently decompose these words into a representation similar to that of a full PP mit das (with that), fr das (for that). This analysis is of little impact at the moment, but could pave the way for determining the referent of the pronominal part. For the purposes of Amalgam, we simply reconstruct the string of the proforms from the representation at LF, undoing the analysis step. Insertion of relative pronouns Relative pronouns in our logical form analyses have been replaced by a copy of the semantic node they are referring to. In order to produce correct output in generation, this copy has to be replaced by a relative pronoun. For an illustration of this, see  REF _Ref530883055 \h Figure 9 and  REF _Ref530884156 \h Figure 10 above. All the relevant information for this replacement is present during the flesh-out stage: we need to know that the node in question is a copy of a node in the parent chain (this information is encoded in the CoIndex attribute), we need to know that the node in question is inside a relative clause (the label RELCL must have been assigned to the parent), and we need to know that the RELCL in question modifies the original node that the copy was made from (this information is directly encoded as a semantic relation in LF). Insertion of reflexive pronouns Reflexive pronouns are used in two contexts in German: there are inherently reflexive verbs, where the reflexive pronoun does not carry any semantic role, and there are normal transitive verbs used reflexively. In the first context, the reflexive does not appear as a node in logical form at all (but the verb is marked with a special feature ReflexSens), in the second context, it appears as a copy of the node that it refers to. Insertion of reflexive pronouns picks up on these two different contexts and inserts a reflexive pronoun in the first context, and replaces the copy with a reflexive pronoun in the second context. Insertion of wie Wie is a Wh adverb, like its English counterpart how. It is not represented as a node at logical form, since its only function is to carry the Wh feature. Insertion of wie is a simple operation that is triggered if there is a Wh feature on a node, but no other Wh-carrying element is present or has been inserted yet. Converting the fleshed-out LF to a basic tree This is a recursive transformation of the degraphed LF tree into a syntactic tree structure, such that the LF node label becomes the head of a constituent, and LF modifier nodes become syntactic modifiers. The syntactic labels that have been assigned during the flesh-out stage are copied onto the corresponding nodes in the syntactic tree (by the function adjust_labels). The splitting of separable prefixes Splitting a verb into a stem and a separable prefix is triggered when the verb is actually a separable prefix verb (as indicated by a lexical feature ) and the verb occurs in a context where the stem should be separated, i.e., either in a verb-initial or in a verb-second structure with no auxiliary or modal verb present that would carry the finiteness features. If these conditions hold, lexical information on the verb determines where the split between stem and prefix should be made. The node is split into a STEM and a PREFIX node (see  REF _Ref530885349 \h Figure 11 for illustration). Verbal inflectional features are copied over to the stem. Introduction of coordination Coordination, one of the notoriously difficult aspects of natural language, is represented in different ways at the logical form level and during syntactic analysis. Syntactically, we treat a conjunction as the head of a coordinated construction, with the coordinated phrases and additional conjunctions in the pre- and postmodifiers of that head. Semantically, there is no single node for the coordinated phrase (see  REF _Ref531761766 \h Figure 19 for an example). Rather, each of the coordinated phrases has its own node, and enters into semantic relations by itself. In addition, each of the coordinated nodes maintains pointers to the semantic nodes of the other phrases with which it is coordinated in an attribute CoCoords. This mismatch in representation is remedied by two functions which adapt the syntactic tree structure that has been built directly from the degraphed logical form. In essence, the functions convert CoCoords into coordinated syntactic nodes, with the conjunction as the head. Rule-based movement operations Raising, Wh movement and movement of relative pronouns/expressions are handled in a rule-based manner in Amalgam, as opposed to the extraposition of clauses, which employs a decision tree classifier. This is by no means a necessary design feature, but given the sparsity of data and the simplicity of a rule-based approach, we decided to adopt this strategy in our prototype. We are well aware of the complexity of long distance movement phenomena across languages, and we do not suggest that the simplistic treatment we have chosen is a linguistically adequate solution. However, it is also a fact that long-distance movement is a rare phenomenon in the data that we currently work with, so that we have no basis in the data for a machine learned approach. There are currently two raising functions. One function raises nodes out of adjective phrases and noun phrases to the level of the copular verb in predicative contexts, the other function raises subjects of raising verbs. The latter function is a prototype that hasn't been tested, since in the data we don't find enough instances of raising verbs to learn the correct syntactic labels for them. Without correct labeling, this function is not triggered. Wh movement is triggered if the structure contains a phrase marked by the Wh feature that is not dominated by another Wh or WhQ phrase (a direct or indirect Wh question) and if that phrase has an ancestor higher up in the tree that is marked as WhQ. Once this context is detected, the Wh phrase is moved up to the WhQ node. Relativizer movement works very similarly to Wh movement, except that the triggering context here is the presence of a relative pronoun that is not dominated by a relative clause. In this context, the relative pronoun moves up to the first relative clause in its parent chain. Placement of inflectional features on verbs Two functions distribute inflectional features to the correct verbal targets. The first function identifies the finite verb (which can be an inserted auxiliary or a modal) and shifts the tense, mood, and finiteness features to that verb. It also marks the non-auxiliary verb as past participle if the construction is marked as perfective or passive, and it marks verbs as infinitives if a modal verb is present and there is no passive or perfective context. The second function identifies the grammatical subject as the first nominative noun phrase that is in the domain of the verb. It then copies person and number features of that noun phrase onto the finite verb. If no grammatical subject is found, a default assignment of 3rd person singular is made. Fixing up surface order Given the current limitation of the generative language model employed for ordering, we employ a fix up function which deals with the current inability of the order model to account reliably for the position of the verb. This function adjusts the verb positions according to the verb position features that have been assigned by a decision tree classifier in the Flesh-out stage. The function shifts the finite verb to the left bracket position in verb-second and verb-initial structures, and ensures that all non-finite verbs are lined up in the right bracket position. Needless to say, we consider this fix up function a temporary solution that will become obsolete as work on the order model progresses. Compound generation Compound nouns are very common in German. In order to provide fluent output, Amalgam needs to be able to provide compounded nouns where the input Logical Form contains a noun modifying another noun directly (i.e. without the presence of a preposition). This scenario exists especially in the context of machine translation, where input Logical Forms are created from Logical Forms of another language such as English. Compound generation in Amalgam is rule-based: nominal modifiers on a noun are strung together into one single word string. Information about the linking morpheme (letters that are inserted between nominal parts, depending on lexical information on the left-hand word) is retrieved from the lexicon and taken into account. If no linker information is available in the lexicon, a hyphen is inserted between the parts of the compound as a back-off strategy to facilitate intelligibility. Inflectional generation For inflectional generation, the terminal nodes in the syntactic tree with their inflectional bits and case information (on nouns) are passed into Nlpwins generation function, which has been developed for the Office grammar checker projects. This generation function utilizes Nlpwin's finite-state morphology, developed with the other analysis components of Nlpwin. The machine-learned components of Amalgam We provide in-depth descriptions of each of the machine-learned components, beginning with the decision tree classifiers and proceeding to the generative language model of syntactic constituent structure. Decision Tree Classifiers All of the machine-learned modules in Amalgam, with the notable exception of the order model, are based on decision tree classifiers. We use the WinMine toolkit (Chickering, nd.) to build and view our decision trees. For each classification task, we build decision trees at varying levels of granularity (by manipulating the prior probability of tree structures to favor simpler structures) and selected the model with maximal accuracy on the corresponding parameter tuning data set.. Reported accuracy numbers are based on that selected model. In this section, we discuss each of the decision tree classifiers in turn, providing information on the motivation for the classifier, the input features, and the features actually selected by WinMine. We evaluate the accuracy of each classifier and perform failure analysis. If not otherwise specified, the decision tree classifiers are built on a set of 100K sentences from the technical domain (computer manuals). The set is split 70/30 for training versus prameter tuning, respectively. We report overall accuracy and baseline accuracy of each model, as well as precision/recall/F-measure for each value of the target feature on the parameter tuning set. The baseline accuracy number is the accuracy resulting from applying the most frequent value of the target feature across the board. Work on the individual models is ongoing, and the numbers reported in this report are from December 2001. Considerable improvements have been made since then. For more recent results, see (Corston-Oliver et al. 2002). We use standardized sets of features for the training of the classifiers, with special features used sparingly and only with linguistic motivation. To simplify the discussion of the features in the individual sections on decision trees, we will use the following set of terms: Standard bits: An inclusive set of features present on records in NLPWIN. These include subcategorization features, semantic features, tense features, Wh etc. The total number of features in this set is 193. The value of these features is binary: either the bit is present, or it is not. Standard attributes: An inclusive set of attributes present at logical form. The total number of attributes is 33. The value of these features is binary: either the attribute is present, or it is not. ParentAttrs: An attribute denoting the semantic relation of a node to its Parent. The value of this feature is an atom, there are as many values as there are logical form standard attributes. Cat: The part of speech Crds and CoCoords: Coordination-related attributes. CoCoords is a list with references to the other nodes with which a semantic node is coordinated. Crds has the same function, but is used in root node contexts. The following is a list of those standard bits that ended up being used in any of the decision tree classifiers: Person/number bits: Sing, Plur, Pers3, Pers2 Verb-related bits: Pass (passive), Pres (present), Indicat (indicative), Futr (future), Perf (perfective), Imper (imperative), Condition (conditional clause), Modal (modal verb), Continuous (aspectual feature), Completed (aspectual feature), Resultat (aspectual feature), Possibl (modal feature), YNQ (yes/no question), WhQ (Wh question), Etreaux (verb takes sein to form the perfect tense), I3 (intransitive verb with infinitival clause), Noun-related bits: N_ung (noun derived from verb by suffix ung), Fem, Masc, Neut, Def (definite), Indef (indefinite), Univ (universal quantification), Proxl (demonstrative, indicating something near to the speaker), Quant (quantified), CompPart (part of a compound), ExstQuant (existential quantification), Wh, Reflex (reflexive), Rel (relative), Subcategorization bits: I0 (intransitive), T1 (transitive), T5 (transitive with that-clause), T1dat (transitive with dative object), T1acc (transitive with accusative object), T1gen (transitive with genitive object), D1 (ditransitive verb), D5 (ditransitive verb with that-clause), L1 (copular verb with NP predicate), Extrap3 (with infinitival complement that can be extraposed), V3 (takes an NP and an infinitival complement), B3 (adjectives/adverbs that take infinitival complements), F5 (adjectives/adverbs that take that-clause complements), V2 (takes a bare infinitive as a complement), T6 (transitive, takes a Wh-clause as complement), T3 (transitive, takes an infinitival clause as complement), V2comp (takes a verb-second complement clause) Miscellaneous bits: Neg (negated), Proposition (has propositional content), Time (time expression) Standard attributes: PrpCnjLem (Lemma of prepositional element), Classifier (classifier expression), LOps (operators), Modals, DegreeMods (modifiers of degree), Intnsifs (intensifiers), Cause, Locn (location), Props (propositions), SMods (sentence modifiers), Time, Manner, Equiv (equivalence relation), Measure, Tsub (subject), Tobj (object), Tind (indirect object), Benef (beneficient), Matr (material), Duration, Possr (possessor), Mod (unspecified modifier), Part (part relation), Means, Purp (purpose), Result, Source, Goal, Attrib (attributive relation), LAgent (agent), CoAgent, PrepRel (unspecified prepositional relation), Appostn (apposition) A special notation is used in the composition of the feature names to indicate if the feature is tested on the Parents, or Parents of the Parents, and in the case of special features, to indicate whether the feature value is an atom or a continuous value. To give some examples: 1~I0 = the I0 bit on the node itself 1~I0~Parents = the I0 bit on the first of the Parents (technically, the Parents attribute is a list, but in a degraphed LF there is (by definition) only one element in that list 1~I0~Parents~Parents = the I0 bit on the first of the Parents of the first of the Parents (i.e., on the grandparent) A~myfeature = a special feature myfeature that has an atom as its value The 1~ prefix simply indicates that only one record is examined to compute this feature. Since this is the case for all features in Amalgam, it can be ignored. To give some ballpark figure for the total number of features that are emitted for each of the classification tasks, there are 193 standard bits in addition to the 33 standard attributes. For those models where we test the standard bits and attributes on a node itself, its parent and its grandparent, we emit a total of (193 + 33) * 3 = 678 features. All the models in the flesh-out stage are trained on (and applied to) logical form nodes, all the models after flesh-out are trained on (and applied to) syntactic nodes in the basic tree. If logical form information needs to be accessed from the basic tree, it is accessed through the SemNode attribute which refers to the corresponding semantic node. Syntactic labeling Motivation In Amalgam, sentence realization is mediated through a syntactic stage. Logical form is converted step by step into a syntactic structure, with the output being very similar to an analysis tree structure. Syntactic labels (especially on non-terminal nodes) are an important part of any syntactic tree, and many linguistic phenomena can be best described with reference to syntactic labels. Input features Cat and ParentAttrs on the node itself, the parent, and the grandparent Standard bits and attributes on the node itself, its parent, and grandparent Two special features: HasWhDaughter: 1 if the node has a daughter that is marked +Wh, 0 otherwise Pers2~Tsub: the Tsub is Pers2 Features selected A total of 48 features were selected for this model: 1~Cat, 1~ParentAttrs, 1~PrpCnjLem, 1~Pers3~Parents, 1~Pass~Parents, 1~LOps, 1~ParentAttrs~Parents~Parents, 1~Continuous~Parents, 1~Time~Parents, 1~Pres~Parents, 1~Resultat~Parents, 1~Pers3, 1~CoCoords, 1~Indicat, 1~Cat~Parents, 1~Tsub~Parents, 1~Proposition, 1~Tsub, 1~ParentAttrs~Parents, 1~T1~Parents, 1~LOps~Parents~Parents, 1~Pres, 1~Tobj, 1~Quant~Parents, 1~I0~Parents, 1~CnjLem, 1~Quant, 1~Def, 1~CoCoords~Parents~Parents, 1~PrpCnjLem~Parents, 1~Crds, 1~CoCoords~Parents, 1~Condition, 1~Indef, 1~Pers2~Tsub, 1~Rel, 1~Plur, 1~T1~Parents~Parents, 1~Attrib, 1~Resultat, A~HasWhDaughter, 1~Sing, 1~Completed, 1~Modal~Parents, 1~Proposition~Parents, 1~Quant~Parents~Parents, 1~T5~Parents, 1~Plur~Parents~Parents Classifier accuracy and complexity This classifier is built on 10,000 sentences (with a 70/30 split training versus parameter tuning/test). Since the classifier is trained on each semantic node that has a corresponding syntactic node, the number of data points obtained from 10,000 sentences is already very large. The accuracy for this model is 98.27%. The baseline for the model is .35. Precision, recall, and F-measure for each of the values of the target feature are given in  REF _Ref531402157 \h Figure 21. The DT classifier has 121 branching nodes. KeyPrecisionRecallF-measureDETP0.9929(140/141)0.9396(140/149)0.9655COMPCL0.9276(141/152)0.8545(141/165)0.8896VP0.9807( 1928/ 1966)0.9954( 1928/ 1937)0.9880QUANP0.9856(618/627)0.9968(618/620)0.9912AVPNP0.0000(0/0)0.0000(0/ 12)0.0000IMPR0.9835(298/303)1.0000(298/298)0.9917AVP0.9982( 1658/ 1661)0.9846( 1658/ 1684)0.9913LABEL0.9296( 66/ 71)0.9041( 66/ 73)0.9167NAPPOS0.9838(426/433)0.9660(426/441)0.9748QUES0.0000(0/0)0.0000(0/5)0.0000AUXP0.9902(906/915)1.0000(906/906)0.9951NREL0.7143( 30/ 42)0.5882( 30/ 51)0.6452AJP0.9909( 3582/ 3615)0.9942( 3582/ 3603)0.9925ABBCL0.0000(0/0)0.0000(0/8)0.0000RELCL0.9871(686/695)0.9985(686/687)0.9928NP0.9755(10906/11180)0.9957(10906/10953)0.9855POSS0.9734( 1535/ 1577)0.9672( 1535/ 1587)0.9703PRPRTCL0.0000(0/0)0.0000(0/2)0.0000COMMENT0.2500(6/ 24)0.8571(6/7)0.3871INFCL0.9007(136/151)0.9645(136/141)0.9315PP0.9923( 5791/ 5836)0.9585( 5791/ 6042)0.9751SUBCL0.9915(585/590)0.9701(585/603)0.9807DECL0.9959( 1695/ 1702)0.9953( 1695/ 1703)0.9956PTPRTCL0.0000(0/0)0.0000(0/4)0.0000Figure  SEQ Figure \* ARABIC 21: Precision, recall, and F-measure for the syntactic label model Failure analysis As is apparent from  REF _Ref531402157 \h Figure 21, a few of the syntactic labels used in Nlpwin are very rare in the training data used, so that the model is not able to correctly pick up on the determining factors for these labels. The problematic labels include: AVPNP (noun phrase used adverbially), QUES (question), ABBCL (absolute clause), PRPRTCL (present participle clause), and PTPRTCL (past participle clause). Determiner insertion Motivation Function words that are resolved as features at the level of logical form need to be re-inserted during sentence realization. Input features Nodetype on the node itself, parent and grandparent Cat, ParentAttrs on the node itself and the parent Cat of the possessor of the node itself Standard bits and attributes on the node itself and the parent Features selected Fourteen features were selected for this model: 1~Def, 1~Nodetype, 1~Indef, 1~Wh, 1~Proxl, 1~Cat~Possr, 1~Nodetype~Parent, 1~Plur, 1~PrpCnjLem, 1~Sing, 1~Quant, 1~Cat~Parents, 1~Resultat~Parents, 1~Cat Classifier accuracy and complexity The classifier accuracy is 97.63%. The baseline is 0.58.  REF _Ref531403222 \h Figure 22 shows the numbers for each of the five observed values of the target feature. The determiner insertion model has nineteen branching nodes. KeyPrecisionRecallF-measureNoDet0.9905 (11284/11392)0.9719 (11284/11610)0.9811Whdet1.0000 (22/22)0.9565 (22/23)0.9778Proxldet0.9922 (508/512)0.9203 (508/552)0.9549DefDet0.9433 (6068/6433)0.9859 (6068/6155)0.9641Indefdet1.0000 (1765/1765)0.9893 (1765/1784)0.9946Figure  SEQ Figure \* ARABIC 22: Precision, recall, and F-measure for the determiner insertion model Failure analysis The majority of incorrect classifications in this model stems from coordinated noun phrases, where (at least for German) the determiner is often only spelled out once. Feature extraction for this model should be refined, so that either coordination is taken into account, or coordinated NPs are consistently ignored during feature extraction. Auxiliary insertion Motivation Function words that are resolved as features at the level of logical form need to be re-inserted during sentence realization. Input features ParentAttrs and Cat on the node itself Standard attributes and bits on the node itself and the parent Crds and CoCoords on the node itself and the parent Standard bits on the LexNode of the node itself Etreaux bit on the LexNode of the node itself Features selected Thirteen features are selected. 1~Pass, 1~Perf, 1~Completed, 1~Proposition, 1~Pres, 1~Condition, 1~CnjLem, 1~Etreaux~LexNode, 1~Past, 1~D1~LexNode, 1~T1, 1~ParentAttrs, 1~Tsub Classifier accuracy and complexity The accuracy of this classifier is 99.86%. The baseline is 81.36%. The classifier has 14 branching nodes. KeyPrecisionRecallF-measure sein_werden1.0000( 59/ 59)0.6413( 59/ 92)0.7815sein0.9800(147/150)0.8698(147/169)0.9216 haben0.9713(744/766)0.9960(744/747)0.9835werden0.9967(15223/15274)0.9969(15223/15271)0.9968none0.9993(71048/71098)0.9997(71048/71068)0.9995Figure  SEQ Figure \* ARABIC 23: Precision, recall, and F-measure for the auxiliary insertion model Failure analysis Recall is somewhat poor with the combination of the two auxiliaries sein and werden (in passive perfective). Although we have not performed any detailed failure analysis, it seems to be no accident that the combination of sein and werden is also the rarest in the data. Preposition insertion Motivation In German, the prepositions von and durch in passive contexts are semantically vacuous and are not represented at the level of logical form. During sentence realization, they need to be inserted under the appropriate circumstances. Input features Standard bits and attributes on the node itself and the parent Standard bits on the grandparent ParentAttrs and Cat on the node itself Lexical bits (subcategorization and nominal derivational bits) on the LexNode of the parent and the grandparent Features selected Nineteen features were selected. 1~Tsub~Parents, 1~Pass~Parents, 1~N_ung~LexNode~Parents, 1~Continuous~Parents, 1~Cat, 1~Sing, 1~ParentAttrs, 1~T1~Parents, 1~Def, 1~Indef, 1~Proposition, 1~Attrib, 1~Quant, 1~Tobj~Parents, 1~Possr, 1~Tobj, 1~Plur~Parents~Parents, 1~Proposition~Parents, 1~Pers2 Classifier accuracy and complexity Accuracy is 99.15%. The baseline was 97.04%. The model has 20 branching nodes. This model was trained on 10k sentences, since each nominal node was taken into consideration, which yielded a large number of data points. KeyPrecisionRecallF-measure von0.8559(202/236)0.7953(202/254)0.8245none0.9951(10502/10554)0.9996(10502/10506)0.9973 durch0.7500( 27/ 36)0.4091( 27/ 66)0.5294Figure  SEQ Figure \* ARABIC 24: Precision, recall and F-measure of the preposition insertion model Failure analysis It is not surprising that it is not easy for the model to predict the choice of prepositions von and durch. The choice of prepositions in the German passive is governed by intricate semantic details in interpretation, which generally go beyond the level of granularity of our logical form representation. A cursory failure analysis on the model confirmed that the most important factor in mis-classifications is indeed the distinction between those two prepositions. Insertion of infinitival marker Motivation As with other function words like auxiliaries and determiners, the infinitival marker is semantically vacuous and therefore is not represented as a semantic node at the level of logical form. During sentence realization, it needs to be inserted under the appropriate circumstances. Input features ParentAttrs, Cat and Nodetype on the node itself Standard bits and attributes on the node itself and on the parent Crds and CoCoords on the node itself and the parent Standard bits on the grandparent Features selected Fourteen of the input features were selected: 1~Nodetype, 1~Pres, 1~PrpCnjLem, 1~I0, 1~Tobj, 1~Tsub~Parents, 1~Modal, 1~ParentAttrs, 1~V2~Parents, 1~PrpCnjLem~Parents, 1~CnjLem, 1~Past, 1~T1, 1~CnjLem~Parents Classifier accuracy and complexity The accuracy of the classifier is 99.77%. The baseline was 95.66%. The decision tree has 15 branching nodes. The model was trained on 10k sentences since every verbal node was a data point. KeyPrecisionRecallF-measurezu0.9634(342/355)0.9856(342/347)0.9744none0.9993( 7630/ 7635)0.9983( 7630/ 7643)0.9988Figure  SEQ Figure \* ARABIC 25: Precision, recall, and F-measure for the classifier for insertion of infinitival markers Failure analysis Cursory failure inspection reveals that faulty parses (yielding erroneous LFs) are an important factor in mis-classifications. Once the overall analysis of the sentence is incorrect, proper identification and marking of infinitivals becomes very difficult. Negation insertion Motivation Negation is represented at the level of logical form as a feature Neg. The decision as to where to insert negation during sentence realization is not a completely straightforward one, though, because during analysis, the Neg feature can percolate up in the tree under various circumstances. Input features ParentAttrs on the node itself Cat and Nodetype on the parent and grandparent Standard attributes and bits on the node itself and the parent Standard bits on the grandparent Crds and CoCoords on the node itself, parent, and grandparent Two special features: Negquant: is 1 if any of the descendants has the ExstQuant bit, indicating that it is in the scope of a negative operator NegquantSister: is 1 if any of the sister nodes has a descendant which bears the ExstQuant bit Features selected Fifty-eight features were selected: 1~Cat,1~Neg, 1~Nodetype, 1~ParentAttrs, F~NegquantSister, 1~Nodetype~Parents, F~Negquant, 1~Sing, 1~Mod, 1~Pass~Parents, 1~Quant, 1~PrepRel, 1~Tobj, 1~Cat~Parents, 1~Nodetype~Parents~Parents, 1~BndPrp, 1~Tsub, 1~Time, 1~Mod~Parents, 1~SMods, 1~Pass, 1~CompPart, 1~Pres, 1~Cat~Parents~Parents, 1~Modals~Parents, 1~Plur, 1~Pers3, 1~ExstQuant, 1~Proxl, 1~Resultat~Parents, 1~Modal, 1~PrepRel~Parents, 1~LAgent~Parents, 1~Proposition, 1~Tsub~Parents, 1~I0, 1~Plur~Parents, 1~Sing~Parents, 1~Indef, 1~CoCoords~Parents, 1~Tobj~Parents, 1~I0~Parents, 1~Def, 1~Condition, 1~Indicat~Parents, 1~Pres~Parents, 1~Univ, 1~Indicat, 1~T1~Parents~Parents, 1~Proposition~Parents, 1~Condition~Parents, 1~Sing~Parents~Parents, 1~Modal~Parents, 1~Def~Parents, 1~Plur~Parents~Parents, 1~Modals, 1~Indef~Parents, 1~Quant~Parents Classifier accuracy and complexity The accuracy is 90.94% with a baseline of 80.79%. The resulting model has 138 branching nodes. KeyPrecisionRecallF-measureInsert_neg0.7969( 2648/ 3323)0.7058( 2648/ 3752)0.7486none0.9319(15100/16204)0.9572(15100/15775)0.9444Figure  SEQ Figure \* ARABIC 26: Precision, recall, and F-measure for insertion of negation Failure analysis It is not surprising that for each given node, the decision of whether negation should be overtly realized at that very position is far from trivial. Since we know from the properties of our logical forms that each Neg feature corresponds to an overtly realized negation at that node or any descendant of that node, we apply the negation insertion model in the following way: If a Neg feature is encountered at node X, a probability for insertion of negation is assigned by the model for X and all its descendants. Whichever node receives the highest probability for insertion will be the target for insertion. Insertion of subordinating conjunctions Motivation The subordinating conjunctions dass and ob are not represented at logical form as nodes, and hence need to be inserted during sentence realization. Input features Standard attributes and bits on the node itself, the parent, and the grandparent ParentAttrs, Cat and Nodetype on the node itself Subcategorization bits on the LexNode of the parent and the grandparent Features selected Twenty-two features were selected: 1~Proposition,1~ParentAttrs,1~Pres~Parents~Parents,1~Pers3~Parents~Parents,1~Proposition~Parents~Parents,1~Equiv,1~Proposition~Parents,1~Pers3~Parents,1~T1acc~LexNode~Parents,1~V2comp~LexNode~Parents,1~Mod,1~Tobj,1~Pres~Parents,1~Pass~Parents,1~D5~Parents,1~T1~Parents,1~L1,1~Pass,1~I0~Parents,1~T1, 1~T1acc~LexNode~Parents~Parents, 1~Plur~Parents Classifier accuracy and complexity The accuracy of the classifier for the insertion of subordinating conjunctions is 95.47%. The baseline is 54.55%. The model contains 27 branching nodes. KeyPrecisionRecallF-measuredass 0.95( 1251/ 1323) 0.92( 1251/ 1353) 0.93none 0.95( 2003/ 2106) 0.97( 2003/ 2075) 0.96ob 1.00(432/432) 1.00(432/433) 1.00Figure  SEQ Figure \* ARABIC 27: Precision, recall and F-measure for the classifier for insertion of subordinating conjunctions Failure analysis It is to be expected that the insertion of ob is very reliable, given the nature of our logical form representation, where whether type clauses are marked with the YNQ feature. The insertion of dass is less straightforward in German: a subset of verbs which take complement clauses actually allow complementizer-less complement clauses (with the same verb position as main clauses). For these verbs, then, there is a genuine choice between a complement clause with dass and one without. The numbers in  REF _Ref531421544 \h Figure 27 bear out that prediction. Insertion of expletive subjects Motivation Expletive (or pleonastic) subjects are semantically empty subjects that are necessary on purely grammatical grounds. They serve as placeholders for other constituents that have been displaced from the subject position. In some constructions, the insertion of expletive subjects is the only option, for example in the existential construction in German: es gibt verschiedene Mglichkeiten (there are different possibilities), which cannot be expressed grammatically without the presence of the es subject. By definition, a semantic representation such a the logical form in Nlpwin will not encode purely grammatical markers such as the expletive subject es in German. This makes it necessary to decide where to insert expletive subjects during sentence realization. Input features ParentAttrs, Cat and Nodetype on the node itself Standard bits and attributes on the node itself, and the parent Standard bits on the grandparent Standard bits and attributes on the Tobj (semantic object) Subcategorization bits on the LexNode of the node itself, the parent, the grandparent, and the Tobj Special features: Cat of the Tobj Geben: i.e. is the Lemma of the node geben or not (this feature allows the model to zero in on the existential construction) Features selected Twenty-seven features are selected: 1~B3~LexNode~Tobj, geben, 1~Tsub~Tobj, 1~ParentAttrs, 1~Nodetype, 1~I3~LexNode, 1~PrepRel, 1~Indicat, 1~T1acc~LexNode~Parents~Parents, 1~Def~Tobj, 1~T1acc~LexNode, 1~I0~LexNode, 1~F5~LexNode~Tobj, 1~V3~LexNode, 1~T3~LexNode, 1~Extrap3~LexNode~Tobj, 1~Possibl, 1~Pers3~Parents~Parents, 1~Sing~Tobj, 1~T1~LexNode, 1~T1dat~LexNode, 1~Modals, 1~L1~LexNode, 1~PrepRel~Parents, 1~T1acc~LexNode~Parents, 1~Proposition, 1~Props~Tobj, 1~Mod Classifier accuracy and complexity The accuracy is 99.69%, with a baseline of 99.0%. The classifier has 8 branching nodes. KeyPrecisionRecallF-measureno0.9971(79613/79846)0.9998(79613/79632)0.9984yes0.9116(196/215)0.4569(196/429)0.6087Figure  SEQ Figure \* ARABIC 28: Precision, recall and F-measure for the expletive subject insertion model Failure analysis It is obvious from  REF _Ref531488842 \h Figure 28 that the recall for insertion of es with a value of about 0.4569 is currently still problematic. Since the precision is rather high, there is indication that some grammatical contexts for the insertion of es are missed, something that will require more detailed failure analysis. Assignment of probabilities for the spellout of NPs Motivation Semantic nodes at the level of logical form are not always overtly realized in the sentence string. The prototypical examples are the subjects of infinitival clauses, which are logically present and part of the interpretation of the sentence, but are not part of the surface string. In our logical form representation, there is a variety of other scenarios where arguments of predicates are present as semantic nodes (and linked to other semantic nodes), but are not present in the corresponding surface string. Our strategy for dealing with these phenomena is to employ a decision tree classifier which will produce a probability for surface realization for any given (subject or object) node. These probabilities are then stored in an attribute on the node in question. During the conversion to the basic tree, within a set of subject/object nodes that are related to each other, the probability values are examined. All nodes with a probability of overt realization below 0.5 are deleted from the tree, with the safeguard that among each set at least one node must be realized (to avoid truncation). Input features ParentAttrs on the node itself and the parent Cat on the node itself Nodetype of the parent Standard bits on the node itself, the parent, and the grandparent Standard attributes on the parent Features selected Fifty-six features are selected for this model: 1~Nodetype~Parents, 1~Sing, 1~ParentAttrs~Parents, 1~Pres~Parents, 1~Cat, 1~Pers2, 1~Plur, 1~Def, 1~Proxl, 1~Pass~Parents, 1~Indef, 1~Proposition~Parents, 1~I0~Parents, 1~CompPart, 1~Quant, 1~Pers3, 1~T1~Parents, 1~Imper~Parents, 1~Continuous~Parents, 1~Modal~Parents, 1~Plur~Parents~Parents, 1~T5~Parents, 1~T1~Parents~Parents, 1~Sing~Parents~Parents, 1~Condition~Parents, 1~Indicat~Parents, 1~Indef~Parents~Parents, 1~Perf~Parents, 1~BndPrp~Parents~Parents, 1~Proposition, 1~Def~Parents, 1~L1~Parents~Parents, 1~L1~Parents, 1~Past~Parents, 1~Pers3~Parents, 1~Imper~Parents~Parents, 1~Pers3~Parents~Parents, 1~Univ, 1~I0~Parents~Parents, 1~CnjLem~Parents, 1~Modal~Parents~Parents, 1~Pass~Parents~Parents, 1~Sing~Parents, 1~Proposition~Parents~Parents, 1~Def~Parents~Parents, 1~D1~Parents~Parents, 1~Indicat~Parents~Parents, 1~Condition~Parents~Parents, 1~V2~Parents~Parents, 1~Pres~Parents~Parents, 1~CompPart~Parents, 1~T3~Parents~Parents, 1~CnjLem~Parents~Parents, 1~Futr~Parents, 1~Plur~Parents, 1~V3~Parents~Parents Classifier accuracy and complexity The accuracy is 88.59%. The baseline is 68.19%. There are 447 branching nodes in the decision tree, making it the most complex classifier in Amalgam. KeyPrecisionRecallF-measureno0.8951 (23473/26224)0.7263 (23473/32319)0.8019yes0.8827 (66540/75386)0.9603 (66540/69291)0.9198Figure  SEQ Figure \* ARABIC 29: Precision, recall and F-measure for the subject/object realization model Failure analysis This model is one of the most complex in Amalgam, indicating that the task at hand is very difficult. While the precision and recall numbers are satisfactory, there is still room for improvement. One possible strategy is to train the model on non-fitted parses only (i.e. only on those parses that have a spanning analysis for the whole input string), to have it focus on true grammatical generalizations, and not be distracted by faulty parses and noise in the input. Assignment of Case Motivation Case is an important feature in the German grammar. Recall from Section  REF _Ref10887901 \r 3.3 that there are four different cases in German (accusative, nominative, dative, and genitive). Constituent order is relatively free in German, and often only the case-marking on a noun phrase will indicate whether it is to be interpreted as the subject, object, or indirect object in a sentence. During sentence realization, case serves as a proxy for grammatical subjecthood etc. For surface realization it is therefore imperative to identify the case of a given noun phrase properly, in order to produce intelligible output. Input features ParentAttrs on the node itself, the parent and the grandparent Nodetype of the node itself, the parent and the grandparent Pred (logical form equivalent of lemma) on the parent Cat on the node itself, the parent and grandparent Standard bits on the node itself (with the exception of person and number bits) Standard bits on the parent and grandparent Standard attributes on the node itself, the parent and the grandparent Lexical subcategorization bits on the LexNode (containing the information after lexical lookup) of the parent and the grandparent Five special features: Parent_lemma_geben: 1 if the lemma of the parent is geben Prep_lemma: the lemma of the governing preposition PrepandPrepdat: 1 if there is a governing preposition and it is marked as Prepdat (governs the dative) PrepandPrepacc: 1 if there is a governing preposition and it is marked as Prepacc (governs the accusative) PrepandPrepgen: 1 if there is a governing preposition and it is marked as Prepgen (governs the genitive) Features selected Seventy-two features were selected by the model building process: A~PrepandPrepdat, 1~Pred~Parents, 1~ParentAttrs, 1~Nodetype, A~Prep_lemma, A~PrepandPrepacc, 1~Pass~Parents, A~PrepandPrepgen, 1~CoCoords, 1~Nodetype~Parents, 1~Neg, 1~Tobj, 1~T1~Parents, 1~Nodetype~Parents~Parents, 1~Indef, 1~Tind~Parents~Parents, 1~ParentAttrs~Parents, 1~Continuous~Parents, 1~BndPrp, 1~Attrib, 1~Resultat~Parents, 1~T1gen~LexNode~Parents, 1~D1~LexNode~Parents, A~Parent_lemma_geben, 1~X7~Parents~Parents, 1~ParentAttrs~Parents~Parents, 1~I0~Parents, 1~Def, 1~Imper~Parents~Parents, 1~CompPart, 1~Cat~Parents, 1~L1~Parents, 1~Sing~Parents, 1~Tobj~Parents~Parents, 1~T5~Parents, 1~Sing~Parents~Parents, 1~Plur~Parents~Parents, 1~Tind~Parents, 1~CoCoords~Parents~Parents, 1~Quant, 1~Past~Parents, 1~Mod, 1~Indef~Parents, 1~Possr, 1~Cat, 1~T1dat~LexNode~Parents~Parents, 1~Proposition~Parents, 1~LOps 1~Plur~Parents, 1~Proposition, 1~Adjdat~LexNode~Parents, 1~T1~Parents~Parents, 1~Modals~Parents, 1~PrepRel~Parents, 1~I0~Parents~Parents, 1~Tobj~Parents, 1~T1dat~LexNode~Parents, 1~Def~Parents, 1~PrepRel, 1~Completed~Parents, 1~PrpCnjLem~Parents, 1~Pass~Parents~Parents, 1~Cat~Parents~Parents, 1~PrepRel~Parents~Parents, 1~Condition~Parents, 1~T1acc~LexNode~Parents 1~Indicat~Parents, 1~Tsub~Parents~Parents, 1~Pres~Parents, 1~T1acc~LexNode~Parents~Parents, 1~Pers3~Parents~Parents, 1~Pres~Parents~Parents Classifier accuracy and complexity Accuracy is 96.02%. The baseline is 0.46. The model has 226 branching nodes. KeyPrecisionRecallF-measureDat0.9562 (17575/18380)0.9797 (17575/17940)0.9678 Acc0.9257 (5132/5544)0.8783 (5132/5843)0.9014 Gen0.9883 (9950/10068)0.9796 (9950/10157)0.9839 Nom0.9563 (4576/4785)0.9460 (4576/4837)0.9512Figure  SEQ Figure \* ARABIC 30: Precision, recall and F-measure for the case model Failure analysis In German, it is often very difficult during analysis to determine exactly which noun phrases belong together and which do not. Since any number of noun phrases can be strung in a sequence in the middle field of the German sentence, misanalyses are common especially with out-of-vocabulary nouns. Not surprisingly, stranded or misanalyzed noun phrases cannot be reliably assigned case, accounting for many of the errors that the case model makes. Assignment of verb position features Motivation As discussed at length in section  REF _Ref530816124 \r \h 3.1, one of the most important aspects of constituent order in German is the correct positioning of the verb. Our strategy in Amalgam is to use a decision tree classifier to assign features that indicate the verb-positioning pattern in the constituent. Downstream models and functions (such as the order model, the order fix-up functions, syntactic aggregation etc.), can then utilize the information present in these features. Input features ParentAttrs on the node itself Nodetype of the node itself, the parent, and the grandparent Standard bits on the node itself, the parent, and the grandparent Standard attributes on the parent, the parent, and the grandparent Lexical subcategorization bits on the parent, and the grandparent Two special features: A~NTlabelCoordmother: the Nodetype of the parent if the node is coordinated A~EmptySubject: 1 if the subject lemma is _X (indicating an empty, un-controlled subject) Features selected Forty-one features were selected: 1~ParentAttrs, 1~Nodetype, A~EmptySubject, 1~Imper, 1~Tobj, 1~Props~Parents, 1~Proposition, 1~Modals, 1~Indicat, 1~T3, 1~CoCoords, 1~T5~Parents, 1~T5, 1~PrepRel, 1~T6, 1~Tsub, 1~Pass, 1~T1, 1~Mod, 1~I0, 1~PrpCnjLem, 1~Modal, 1~Neg, 1~Proposition~Parents, 1~Condition, 1~Perf, 1~CnjLem, 1~V2comp~LexNode~Parents, 1~L1, 1~Mod~Parents, 1~Pres~Parents, 1~Def~Parents, 1~Tind, 1~Props, 1~Nodetype~Parents, 1~Time, 1~Modal~Parents, 1~WhQ, 1~Modals~Parents, 1~YNQ, 1~Tsub~Parents Classifier accuracy and complexity The accuracy is 94.66%, with a baseline of 0.42. The resulting model has 115 branching nodes. KeyPrecisionRecallF-measure initial0.9650 (7107/7365)0.9818 (7107/7239)0.9733 final0.9374 (16669/17782)0.9750 (16669/17097)0.9558 undefined0.5946 (776/1305)0.3673 (776/2113)0.4541 second0.9721 (18637/19172)0.9719 (18637/19175)0.9720Figure  SEQ Figure \* ARABIC 31: Precision, recall and F-measure of the verb position model Failure analysis It is reassuring that the determination of verb-second, verb-final, and verb-initial patterns can be made at a fairly high level of precision and recall. There are, however, quite a few cases where the verb position is determined to be undefined. A legitimate example of indeterminacy of verb position is a simple construct of the form er geht (he goes). This verb phrase can be either a verb-second verb phrase (as witnessed by the fact that it can stand alone as a declarative sentence), and a verb-final verb phrase (if, for example, preceded by a subordinating conjunction which requires verb-final position as in dass er geht (that he goes). For these undefined cases, the only way to determine the correct verb-position pattern is by linguistic generalization as in this verb phrase is used as a declarative sentence, so in this case it has to be a verb-second structure according to the grammar of German. Since this line of reasoning is not applicable by a purely data-driven model, we will always have a set of data that will escape complete classification into a verb-position pattern. Inversion of dominance Motivation There are grammatical constructions where syntactic dominance is the opposite of semantic dominance. For German as we treat it in NLPWIN, the only example is a quantified expression of the form viele der Leute (many of the people), where viele is the syntactic head of the noun phrase, but semantically it is Leute which is the head, with Nlpwin viele as an operator modifying it. While this phenomenon is rather limited in German, we found that in the NLPWIN analysis of French, modal verbs are treated as syntactic heads (because modals in French pattern syntactically with main verbs rather than with non-modal auxiliary verbs), with a reversal of dominance at the level of logical form. In order to prepare for porting Amalgam to French, we decided to face this issue right away, even though the effect in German is very limited due to data sparsity, and very straightforward since there is only one particular construction which exhibits this property. The approach we have taken is to learn in a decision tree classifier the circumstances under which the reversal of dominance between syntactic and semantic relations takes place, and then perform a reversing operation in the basic tree accordingly. Input features Nodetype and Cat of node itself and parent Standard attributes and bits on the node itself, the parent, and the grandparent Note that the switch model and all following models are trained on, and operate on the basic tree, not the logical form. Only the flesh-out models are applied to logical form structures. Therefore, features of the form 1~X~SemNode mean the presence of feature X on the logical form node (SemNode) of the current node, whereas features of the form 1~X mean the presence of X on the current node. Features selected For German, only two features were selected: 1~Nodetype~SemNode, 1~LOps~Parents~Parents~SemNode Classifier accuracy and complexity The accuracy is 99.69%. The baseline is 92.0%. The model for German is exceedingly simple, with only 3 branching nodes. Key PrecisionRecallF-measure no0.9998( 5974/ 5975)0.9968( 5974/ 5993)0.9983 yes0.9645( 516/ 535)0.9981( 516/ 517)0.9810Figure  SEQ Figure \* ARABIC 32: Precision, recall and F-measure for the dominance switching model Failure analysis As stated above, this model is of an exploratory nature, in preparation for other languages. It is not surprising that for the one simple construction in German that exhibits dominance switching, it is easy to learn the triggering factors reliably. Extraposition Motivation See the detailed discussion of the importance of extraposition phenomena in German in section  REF _Ref531501615 \r \h 3.5 and the discussion in Gamon et al. (2002b). Our approach to extraposition is to determine for each extraposable node (INFCL, COMPCL, RELCL) whether the node should move up one step from its current attachment (its parent node) to the next higher node (its grandparent). From the new position, another assessment is made for the next possible movement step and so on. For training, we extract the value Yes for the target feature for each intermediate node between the source position of a clause and its extraposed position - meaning that the answer to the question should the clause move up one step? is Yes. Similarly, we extract the value No for the extraposed position (since the clause obviously has not moved higher from there). In the case of non-extraposed clauses we extract a No for the parent node of the clause, meaning that there is no movement from the current position. Input features Nodetype of cargo node Nodetype of the potential origination node for movement, of its parent, and its grandparent ParentAttrs of the potential origination node for movement, of its parent, and its grandparent Vfinal feature on the potential origination node for movement, on its parent, and its grandparent Vsecond feature on the potential origination node for movement, on its parent, and its grandparent HasSepfix (indicating a separable prefix verb) on the potential origination node for movement, on its parent, and its grandparent Standard attributes and bits on the the potential origination node for movement, on its parent, and its grandparent Standard attributes and bits on the cargo node (the RELCL/INFCL/COMPCL that could potentially be extraposed), its parent, and grandparent Nodetype of the cargo node ParentAttrs on the cargo node Six special features zeroing in on the relevant aspects of verb position and heaviness as triggers for extraposition: F~NumTokens: number of tokens of the cargo node F~SentenceLengthInToken: length of whole sentence in tokens F~NumChars: number of characters of the cargo node F~SentenceLengthInChar: length of whole sentence in characters A~inVfinalVP: yes if any ancestor of the cargo node is a Vfinal VP, no otherwise A~inVsecondVP: yes if any ancestor of the cargo node is a Vsecond VP, no otherwise Features selected Sixty features were selected during the model building process (features with the 1~ prefix are extracted on the node under consideration for the movement one step up yes or no classification, features with the 2~ prefix are extracted on the extraposable node): 1~Tsub~SemNode, 1~HasSepfix~LexNode~SemNode~Parent, 1~ParentAttrs~SemNode, A~inVfinalVP, 1~Modals~SemNode~Parent, 1~HasSepfix~LexNode~SemNode~Parent~Parent, 1~Pass~SemNode~Parent, 1~Pass~SemNode~Parent~Parent, 1~Modals~SemNode~Parent~Parent, F~SentenceLengthInChar, 1~ParentAttrs~SemNode~Parent~Parent, 2~Pass~SemNode~Parent~Parent, 2~Proposition~SemNode, 1~Nodetype~Parent, 1~Pers3~SemNode~Parent, 1~Vsecond~Parent, 2~Modals~SemNode, 1~Nodetype, 2~T1~SemNode~Parent~Parent, 1~Mod~SemNode~Parent, 1~PrepRel~SemNode~Parent, A~inVsecondVP, 1~PrepRel~SemNode, 1~Tobj~SemNode~Parent, 1~Vfinal, 1~Tsub~SemNode~Parent, 2~Tsub~SemNode~Parent~Parent, 2~Attrib~SemNode~Parent, F~NumChars, 1~BndPrp~SemNode, 1~Nodetype~Parent~Parent, 1~ParentAttrs~SemNode~Parent, 2~Mod~SemNode~Parent~Parent, 2~Indicat~SemNode~Parent~Parent, F~NumTokens, 2~ParentAttrs~SemNode, 2~Nodetype, 1~T1~SemNode~Parent~Parent, 2~Proposition~SemNode~Parent~Parent, 2~CoCoords~SemNode~Parent, 2~Indef~SemNode~Parent 1~Def~SemNode, 1~PrepRel~SemNode~Parent~Parent, 2~PrepRel~SemNode, 1~Mod~SemNode~Parent~Parent, 1~Pers3~SemNode, 1~PrpCnjLem~SemNode~Parent, 1~Plur~SemNode~Parent, 2~PrepRel~SemNode~Parent~Parent, 1~Sing~SemNode~Parent, 2~Def~SemNode~Parent, 1~Plur~SemNode~Parent~Parent, 2~Tobj~SemNode~Parent~Parent, 2~PrpCnjLem~SemNode~Parent, 2~T1~SemNode, 1~Sing~SemNode, 2~I0~SemNode~Parent~Parent, 2~Pers3~SemNode~Parent, 2~Plur~SemNode~Parent, 1~Plur~SemNode Classifier accuracy and complexity The accuracy is 88.16% with a baseline of 0.67. The model has 116 branching nodes. Key PrecisionRecallF-measureNo0.9181 (4720/5141)0.9051 (4720/5215)0.9115Yes0.8094 (2102/2597)0.8331 (2102/2523)0.8211Figure  SEQ Figure \* ARABIC 33: Precision and recall of the extraposition model Failure analysis Extraposition is a non-trivial linguistic phenomenon with a complicated array of triggering factors, including some not completely understood notion of heaviness (where in general a heavier clause tends to extrapose more easily than a lighter clause - see the discussion in Uszkoreit et al 1998). Additionally, the failure to extrapose, or extraposition in a situation where there should not be any often results not in ungrammatical sentences, but in sentences with varying degrees of unnaturalness and lack of fluency. Detailed error analysis in this context would benefit greatly from correlation with human judgements, an experiment that we have not yet undertaken. Realization of determiners Motivation There are 55 different determiner forms observed in the training data. While the realization of the determiner could be determined by rule, we decided to train a decision tree qualifier for the task. Input features Lemma ParentAttrs on the SemNode Gender and number bits on the parent Case on the parent Standard attributes on the parent Features selected Eighteen features were selected. F~ParentCase, 1~ParentAttrs~SemNode~Parent, 1~Lemma, 1~Fem~Parent, 1~Masc~Parent, 1~Neut~Parent, 1~Plur~Parent, 1~ParentAttrs~SemNode~Parent~Parent, 1~Sing~Parent, 1~Mod~SemNode~Parent, 1~Tobj~SemNode~Parent, 1~LOps~SemNode~Parent, 1~PrepRel~SemNode~Parent, 1~PrpCnjLem~SemNode~Parent, 1~Attrib~SemNode~Parent, 1~Appostn~SemNode~Parent, 1~Possr~SemNode~Parent, 1~Tsub~SemNode~Parent Classifier accuracy and complexity The accuracy is 90.77% with a baseline of 21.65%. The resulting decision tree classifier has 266 branching nodes. KeyPrecisionRecallF-measure derjenigen0.0000( 0/ 0)0.0000( 0/ 3)0.0000 denselben0.0000( 0/ 0)0.0000( 0/ 45)0.0000 meiner0.0000( 0/ 0)0.0000( 0/ 1)0.0000 dieselben0.0000( 0/ 0)0.0000( 0/ 40)0.0000 ein_und_derselbe0.0000( 0/ 0)0.0000( 0/ 1)0.0000 meines0.0000( 0/ 0)0.0000( 0/ 2)0.0000 desselben0.0000( 0/ 0)0.0000( 0/ 13)0.0000 sein0.3333( 12/ 36)0.8000( 12/ 15)0.4706 derselbe0.0000( 0/ 0)0.0000( 0/ 4)0.0000 das0.9148( 3531/ 3860)0.8768( 3531/ 4027)0.8954 die0.9432( 13327/ 14129)0.9091( 13327/ 14659)0.9259 dem0.9003( 7402/ 8222)0.9426( 7402/ 7853)0.9209 den0.8894( 4320/ 4857)0.8171( 4320/ 5287)0.8517 einem0.8134( 1539/ 1892)0.8927( 1539/ 1724)0.8512 einen0.8409( 1438/ 1710)0.8489( 1438/ 1694)0.8449 der0.9042( 14292/ 15807)0.9504( 14292/ 15038)0.9267 des0.9771( 3378/ 3457)0.9740( 3378/ 3468)0.9756 dasselbe0.0000( 0/ 0)0.0000( 0/ 17)0.0000 einer0.8931( 1954/ 2188)0.8890( 1954/ 2198)0.8910 eines0.9089( 1067/ 1174)0.9744( 1067/ 1095)0.9405 dies0.0000( 0/ 0)0.0000( 0/ 3)0.0000 diesem0.8551( 655/ 766)0.9590( 655/ 683)0.9041 diesen0.7409( 266/ 359)0.8837( 266/ 301)0.8061 demselben0.0000( 0/ 0)0.0000( 0/ 63)0.0000 denjenigen0.0000( 0/ 0)0.0000( 0/ 2)0.0000 derselben0.0000( 0/ 0)0.0000( 0/ 57)0.0000 meine0.0000( 0/ 0)0.0000( 0/ 7)0.0000 dieser0.7526( 791/ 1051)0.9295( 791/ 851)0.8318 dieses0.9296( 581/ 625)0.9222( 581/ 630)0.9259 diese0.9057( 1248/ 1378)0.8820( 1248/ 1415)0.8937 seinem0.2857( 6/ 21)1.0000( 6/ 6)0.4444 welche0.9118( 124/ 136)0.8671( 124/ 143)0.8889 seinen0.0000( 0/ 0)0.0000( 0/ 12)0.0000 diejenige0.0000( 0/ 0)0.0000( 0/ 1)0.0000 ihr0.8039( 123/ 153)0.8913( 123/ 138)0.8454 seiner0.5313( 17/ 32)0.7083( 17/ 24)0.6071 seines0.0000( 0/ 0)0.0000( 0/ 9)0.0000 ihrem0.8974( 175/ 195)0.8663( 175/ 202)0.8816 ihren0.7818( 86/ 110)0.7748( 86/ 111)0.7783 welchem0.5833( 14/ 24)0.9333( 14/ 15)0.7179 welchen0.0000( 0/ 0)0.0000( 0/ 15)0.0000 ihrer0.8296( 112/ 135)0.8682( 112/ 129)0.8485 mein0.0000( 0/ 0)0.0000( 0/ 1)0.0000 ihres0.8082( 59/ 73)0.9077( 59/ 65)0.8551 welcher0.4627( 31/ 67)0.9118( 31/ 34)0.6139 ihre0.9012( 155/ 172)0.8031( 155/ 193)0.8493 welches0.0000( 0/ 0)0.0000( 0/ 20)0.0000 dieselbe0.0000( 0/ 0)0.0000( 0/ 41)0.0000 ein0.8845( 2511/ 2839)0.9003( 2511/ 2789)0.8923 ein_und_demselben0.0000( 0/ 0)0.0000( 0/ 1)0.0000 diejenigen0.0000( 0/ 0)0.0000( 0/ 7)0.0000 eine0.9606( 3823/ 3980)0.8926( 3823/ 4283)0.9253 ein_und_derselben0.0000( 0/ 0)0.0000( 0/ 1)0.0000 meinem0.0000( 0/ 0)0.0000( 0/ 2)0.0000 seine0.8148( 22/ 27)0.5946( 22/ 37)0.6875Figure  SEQ Figure \* ARABIC 34: Precision, recall and F-measure for determiner realization Failure analysis  REF _Ref531504567 \h Figure 34 shows that not all determiners can be realized with the same accuracy. While the numbers are high for the common determiners such as the definite (der/die/das) and the indefinite determiner (ein/eine), the numbers degrade the rarer the form of the determiner is. It is worth noting that there is a fair amount of linguistic ambiguity: some nouns have more than one gender (stemming from different senses), some nouns can be both singular and plural. Additionally, the level of granularity of the features used to describe the semantic import of determiners at logical form is not sufficient in all cases to uniquely determine the particular lexical choice of determiner. Realization of relative pronouns Motivation Twenty-three different forms of relative pronouns are present in the training set. As with the realization of determiners, we have chosen to build a model for the correct choice, instead of hand-crafting a selection process. Input features Lemma of the grandparent and the great-grandparent Case Nodetype ParentAttrs on the SemNode and on the SemNode of the grandparent Gender and number features on the grandparent and the great-grandparent Standard attributes on the grandparent and great-grandparent Features selected Nineteen variables were selected: F~RelproCase, 1~Masc~Parent~Parent, 1~Fem~Parent~Parent, 1~Plur~Parent~Parent, 1~Nodetype, 1~Neut~Parent~Parent, 1~Sing~Parent~Parent, 1~Lemma~Parent~Parent, 1~ParentAttrs~SemNode, 1~ParentAttrs~SemNode~Parent~Parent, 1~CoCoords~SemNode~Parent~Parent, 1~Fem~Parent~Parent~Parent, 1~Masc~Parent~Parent~Parent, 1~Sing~Parent~Parent~Parent, 1~Possr~SemNode~Parent~Parent, 1~Plur~Parent~Parent~Parent, 1~Lemma~Parent~Parent~Parent, 1~Attrib~SemNode~Parent~Parent~Parent, 1~Tobj~SemNode~Parent~Parent Classifier accuracy and complexity The accuracy is 87.79%. The baseline is 53.59%. There are 77 branching nodes in the decision tree classifier. KeyPrecisionRecallF-measure wozu0.0000( 0/ 0)0.0000( 0/ 1)0.0000 dessen1.0000( 46/ 46)0.8214( 46/ 56)0.9020 die/der0.0000( 0/ 0)0.0000( 0/ 2)0.0000 was1.0000( 10/ 10)0.4167( 10/ 24)0.5882 welche0.0000( 0/ 0)0.0000( 0/ 13)0.0000 wo0.7000( 7/ 10)1.0000( 7/ 7)0.8235 deren0.9406( 95/ 101)0.9896( 95/ 96)0.9645 derer0.0000( 0/ 0)0.0000( 0/ 2)0.0000 das0.9202( 369/ 401)0.7953( 369/ 464)0.8532 dem0.8122( 480/ 591)0.8743( 480/ 549)0.8421 die0.9225( 3248/ 3521)0.9425( 3248/ 3446)0.9324 womit0.0000( 0/ 0)0.0000( 0/ 2)0.0000 den0.8875( 142/ 160)0.6544( 142/ 217)0.7533 der0.8400( 882/ 1050)0.7854( 882/ 1123)0.8118 woraus0.0000( 0/ 0)0.0000( 0/ 1)0.0000 welcher0.0000( 0/ 0)0.0000( 0/ 2)0.0000 welches0.0000( 0/ 0)0.0000( 0/ 1)0.0000 woher0.0000( 0/ 0)0.0000( 0/ 1)0.0000 warum0.0000( 0/ 0)0.0000( 0/ 2)0.0000 wobei0.6907( 67/ 97)1.0000( 67/ 67)0.8171 das/der0.0000( 0/ 0)0.0000( 0/ 1)0.0000 denen0.6727( 298/ 443)0.9113( 298/ 327)0.7740 wodurch0.0000( 0/ 0)0.0000( 0/ 26)0.0000Figure  SEQ Figure \* ARABIC 35: Precision, recall and F-measure for the realization of relative pronouns Failure analysis Similar to the situation with the realization of determiners, it is mostly the rarer forms of relative pronouns that cannot be accurately determined, due to data sparsity. The same caveats with respect to linguistic ambiguity hold for both determiners and relative pronouns. Syntactic aggregation Motivation Any semantic representation of coordination has to encode more than may be present in the syntactic structure. Consider a simple sentence such as John cooked and ate the steak. Semantically, there are two conjoined propositions, John cooked the steak and John ate the steak. There is nothing grammatically wrong with spelling out those two conjoined propositions in the lengthy form John cooked the steak, and John ate the steak. Natural language, however, tends to employ strategies to reduce redundant material in coordination by deleting some of the duplicates. This is generally viewed as a sub-area of aggregation in the generation literature (Wilkinson 1995, Shaw 1998, Reape and Mellish 1999, Dalianis and Hovy 1993). In Amalgam, we only deal with sentence-based realization tasks currently, so the approach we take is strictly intra-sentential, along the lines of what has been called conjunction reduction in the linguistic literature (McCawley 1988). While this may seem a fairly straightforward task compared to inter-sentential, semantic and lexical aggregation, it should be noted that the cross-linguistic complexity of the phenomenon makes it much less trivial than a first glance at English would suggest. In German, for example, position of the verb in the coordinated VPs plays an important role in determining which duplicated constituent can be omitted. In Amalgam, we try to arrive at a reasonable level of fluency in our output, which makes it necessary to model these reduction phenomena in coordination. The model is trained on and applied to the logical form representation that corresponds to the syntax tree. Input features Cat of the node itself, the parent and the grandparent ParentAttrs of the node itself, the parent and the grandparent Nodetype of the node itself, the parent and the grandparent Standard bits and attributes on the node itself, the parent and the grandparent Vsecond and Vfinal features on the node itself, the parent and the grandparent Two special features: F~HeadMod: indicates whether the node in question is a premodifier or a postmodifier of the head of its parent F~AllVerbpos: indicates for VP-coordination if all coordinated VPs are Vfinal or Vsecond Features selected Fifteen features are selected during the construction of the model: A~HeadMod, 1~Proposition~Parents, A~AllVerbpos, 1~Tobj~Parents, 1~Nodetype~Parents, 1~Nodetype, 1~ParentAttrs, 1~ParentAttrs~Parents, 1~CoCoords, 1~T1~Parents, 1~ParentAttrs~Parents~Parents, 1~Cat, 1~Plur~Parents, 1~Proposition~Parents~Parents, 1~CoCoords~Parents~Parents Classifier accuracy and complexity The accuracy is 96.93%, with a baseline of 0.85. The resulting model has 21 branching nodes. The values of the target feature are last, first, and middle: last indicates that the node in question is spelled out in the last of the coordinated constituents, first indicates that it is spelled out in the first of the coordinated constituents, and middle indicates that it is spelled out in a coordinated constituent that is neither first nor last. KeyPrecisionRecallF-measureLast0.9164 (986/1076)0.8851 (986/1114)0.9005First0.9786 (6022/6154)0.9854 (6022/6111)0.9820Middle0.0000 (0/0)0.0000 (0/5)0.0000Figure  SEQ Figure \* ARABIC 36: Precision, recall and F-measure for the syntactic aggregation model Failure analysis The data confirm the initial linguistic hypothesis that coordination reduction is a matter of spelling out a constituent either at the beginning or at the end of coordination, but not somewhere in the middle. Cursory error inspection shows that most of the misclassifications seem to result from either bad analyses, or situations where the verb position has not been uniformly determined as Vsecond or Vfinal in all coordinated VPs. Punctuation Motivation Clearly, punctuation is different from the previously discussed phenomena: it is not a core linguistic phenomenon, but rather a matter of orthographic convention. There are two reasons, however, why we believe that punctuation should be part of Amalgam: without appropriate punctuation, the output of generation - especially for real-life, complex sentences - is difficult to read and hard to parse for a human consumer. even though punctuation is an orthographic convention, it is based on linguistic structure: most punctuation rules make reference to constituenthood, types of constituents etc. Based on the observation that most (if not all) punctuation conventions which we are aware of are of the form insert punctuation mark X before/after Y, but none is of the form insert punctuation mark X between Y and Z, we decided to build two different models for preceding punctuation and following punctuation. At runtime, at each juncture between two words, both models are queried for each non-terminal node in the parent chain. If one of them indicates a high probability of a certain punctuation mark, that vote wins out and the punctuation mark is inserted. Input features The features for the punctuation models are different from the feature sets used in other models. Most of the features are special features, checking the tree configuration: Nodetype and Nodetype of the head Nodetype of the parent and Nodetype of the head of the parent ParentAttrs on the SemNode SentenceLengthInToken and SentenceLengthInChar AtRightEdgeOfParent/AtLeftEdgeOfParent: indicating whether the node is at the right/left edge of its parent node NumTokens and NumChars: number of tokens/chars of the node DistanceToSentenceInitialInToken and DistanceToSentenceFinalInToken DistanceToSentenceInitialInChar and DistanceToSentenceFinalInChar FirstLemma and LastLemma NodetypeOfLeftMostDaughter and NodetypeOfRightMostDaughter NodetypeOfTopRightEdge and NodetypeOfTopLeftEdge: Nodetype of the largest ancestor node with the same right/left edge NodetypeOfLargestPreceedingNT and NodetypeOfSmallestPreceedingNT: Nodetype of the largest/smallest preceding non-terminal node NodetypeOfLargestFollowingNT and NodetypeOfSmallestFollowingNT: Nodetype of the largest/smallest following non-terminal node Features selected In the model for preceding punctuation, all of the features listed above were selected, with the exception of NodetypeOfTopRightEdge. In the model for following punctuation, two features were not selected: DistanceToSentenceFinalInToken and SentenceLength-InToken. Classifier accuracy and complexity The accuracy of the model for preceding punctuation is 98.65%, with a baseline of 89.61%. KeyPrecisionRecallF-measure COMMA0.9500( 29727/ 31293)0.9228( 29727/ 32213)0.9362 OTHERS0.0000( 0/ 0)0.0000( 0/ 23)0.0000 DASH0.0000( 0/ 0)0.0000( 0/ 83)0.0000 SEMICOLON0.0000( 0/ 0)0.0000( 0/ 33)0.0000 NULL0.9907( 280067/ 282705)0.9945( 280067/ 281610)0.9926 COLON0.8153( 203/ 249)0.7123( 203/ 285)0.7603Figure  SEQ Figure \* ARABIC 37: Precision, recall, and F-measure for the preceding punctuation model The accuracy of the model for following punctuation is 98.48% with a baseline of 94.98%. KeyPrecisionRecallF-measure COMMA0.8795( 12462/ 14169)0.8084( 12462/ 15416)0.8425 DASH0.0000( 0/ 0)0.0000( 0/ 45)0.0000 SEMICOLON0.0000( 0/ 0)0.0000( 0/ 27)0.0000 NULL0.9897( 294194/ 297241)0.9943( 294194/ 295884)0.9920 COLON1.0000( 125/ 125)0.7669( 125/ 163)0.8681Figure  SEQ Figure \* ARABIC 38: Precision, recall and F-measure for the following punctuation model Failure analysis  REF _Ref531665805 \h Figure 37 and  REF _Ref531665809 \h Figure 38 show that predictions are reliable for commas, and somewhat reliable for colons. For other punctuation, such as dash and semicolon, data are simply too sparse. The Order Model Motivation Word order plays a crucial role in establishing the fluency and the intelligibility of a sentence. As section  REF _Ref531573279 \n \h 3 explains, word order can make the difference between sensibility and gibberish, especially in a German sentence. Given a syntax tree for a sentence with unordered constituents, such as the tree in  REF _Ref536603665 \h Figure 39, the goal of the Amalgam ordering stage is to establish linear order within each constituent, so that the head and each modifier are placed in their proper position. The ordering stage handles each constituent independently and in isolation, but the net effect is to establish linear order among all leaves of the tree. In our example, the constituent DECL3 has head PREFIX1 (head denoted in the figure by the asterisk *) with modifiers VP2, NP5, NP6, and RELCL3. The ordering stage places these in order independently of the head and modifiers of other constituents, such as RELCL3, for example.  Figure  SEQ Figure \* ARABIC 39: The syntax tree for the sentence Hans isst die Kartoffeln auf, die er gestern geernet hat before ordering  REF _Ref536603733 \h Figure 40 displays the tree after being ordered. The ordering stage has placed the children of DECL3 in the order NP5, VP2, NP6, PREFIX1, and RECLCL3.  Figure  SEQ Figure \* ARABIC 40: The syntax tree after ordering Model and Features The Amalgam ordering stage employs a generative statistical model of syntactic tree structure to score possible orders among a head and its modifiers. The term generative refers to the fact that the distributions in the model could be sampled to generate or build actual syntax trees from scratch, in distribution consistent with the model. It is a useful conceptual framework, even though we are not actually creating the tree at this point in the sentence realization process. For a given constituent, the model assigns a probability to modifier sequences in the context of several relevant features. Many features can be used as relevant context; in practice, our implemented model currently employs the following features: nodetype of the parent of the head (i.e., the constituent type), nodetype of the head (i.e., head part-of-speech) Other possible contextual features include: lemma of the head verb position bits on the parent of the head nodetype of the grandparent of the head presence of an auxiliary in the constituent Given the context features, the model assigns probability to a modifier sequence. Currently each modifier consists of two features: semantic relation (from the logical form) of the modifier to the head nodetype (part-of-speech) of the modifier Other possible features of a modifier include: lemma of the modifier The model is currently constructed to approximate modifier sequence probabilities with an n-gram model. Given a particular context, the model assigns a probability to the semantic relation (from logical form) of each modifier, in the constituents context and in the context of the preceding n-1 neighbors, and it assigns a probability to the nodetype (syntactic category) of the modifier. In the current system, the number of preceding neighbors currently considered is only one; hence, the order model employs a context-dependent bigram. Here is a schematic for a constituent that illustrates the context of parent and head as well as the pre-modifiers and post-modifiers:  EMBED Visio.Drawing.6  Figure  SEQ Figure \* ARABIC 41: Constituent order schematic The model is split into a model of head pre-modifier order (on the left of  REF _Ref536603861 \h Figure 41) and of head post-modifier order (on the right of the figure). Included in the notion of modifier are explicit pseudo-modifiers for marking the beginning and end of the pre-modifiers ( and , respectively) and for marking the endpoints of the post-modifiers ( and ), as shown in the figure. Hence, for any Parent/Head context, the model includes an n-gram distribution for pre-modifiers and an n-gram distribution for post-modifiers. All such distributions are encoded in a single model file.  REF _Ref536603888 \h Figure 42 contains a fragment of a model file for illustrative purposes: [356/563] ( DECL VERB ) Time : AVP [16/563] ( DECL VERB ) Time : AVPNP [10/563] ( DECL VERB ) Time : NP [168/563] ( DECL VERB ) Time : PP [13/563] ( DECL VERB ) Time : SUBCL Figure  SEQ Figure \* ARABIC 42: Model file fragment for the nodetype feature of a pre-modifier with semantic relation Time It shows the context with a DECL (declarative sentence node) as parent and a VERB as head. The fragment shows the distribution for the nodetype feature of a pre-modifier with semantic relation Time preceding the marker (working out from the head). As indicated, such a modifier has probability 356/563 of being an AVP, 16/563 for AVPNP, 10/563 for NP, 168/563 for PP, and 13/563 for SUBCL in that context. Search and Complexity The ordering stage must search among all possible orders or at least among the most promising orders. The search proceeds by considering all possible incomplete orderings of length one, then two, and so on, up to all possible complete orderings of length n. Each step in the search can be pruned to consider only those incomplete order hypotheses for which the model assigns a sufficiently high score. This search is capable of producing as many scored order hypotheses as one cares to retrieve from the final step of the search and is commonly called a beam search, since the threshold for determining a sufficiently high score is often termed the beam. For n members (counting the head and its modifiers), there are n! possible orderings; hence the search space can be overwhelmingly large for a heavy constituent. The beam search constrains the complexity of the complete search and is nearly optimal. Performance Performance of Amalgam was evaluated on a 550MHz PC. On a set of 260 sentences from the technical domain (computer manuals), generation time from logical forms (without the analysis part) was 0.30 seconds per sentence or 3.25 sentences per second. The sentences in the test file had an average length of 15 words per sentence. Evaluation In April 2002, we evaluated the overall system by parsing a blind and randomized test set of 564 German sentences to produce logical forms and then applying Amalgam to generate output strings from those logical forms. For this sample, 71.1% of the words are correctly inflected and occur in the correct position in the sentence. We also compute the word-level string edit distance of the generated output from the original reference string: the number of errors (insertions, deletions, and substitutions) is 44.7% of the number of words in the reference string. String edit distance is a harsh measure of sentence realization accuracy. Because string edit distance does not consider movement as an edit operator, movements appear as both deletions and insertions, yielding a double penalty. Furthermore, as observed earlier, some edits have a greater impact on the intelligibility of the output than others, especially the position of the German verb. Work in progress on a tree edit distance metric addresses both of these issues (Ringger et al., in preparation). Closely related work includes that of Bangalore, Rambow and Whittaker (2000). It is possible that generated sentences might differ from the reference sentences and yet still prove satisfactory. We therefore had five independent human evaluators assess the quality of the output for that same blind and randomized test set of 564 sentences. These sentences had been analyzed to yield logical forms from which Amalgam generated output sentences. The evaluators assigned an integer score to each sentence, comparing it to the reference sentence using the scoring system given in  REF _Ref8111122 \h Table 1. The average score was 2.96 with a standard deviation of 0.81. The mode was 4, occurring 104 times, i.e. 104/564 sentences, or 18.4% received the maximum score. In 63 of these cases, the score of 4 had been automatically assigned because the output sentence was identical to the reference sentence. In the other 41 cases, all five human evaluators had assigned a score of 4, i.e., the output differed from the reference sentence, but was still Ideal. 1. Unacceptable. Absolutely not comprehensible and/or little or no information transferred accurately. 2. Possibly Acceptable. Possibly comprehensible (given enough context and/or time to work it out); some information transferred accurately. 3. Acceptable. Not perfect (stylistically or grammatically odd), but definitely comprehensible, AND with accurate transfer of all important information. 4. Ideal. Not necessarily a perfect translation, but grammatically correct, and with all information accurately transferred. Table  SEQ Table \* ARABIC 1: Evaluation guidelines Using Amalgam in Machine Translation: First Results After we had implemented the German-to-German Amalgam prototype, we started using it in a machine translation context. In machine translation, the logical form representation that serves as input to Amalgam is not produced from the analysis of a German sentence, but is transferred through learned mappings between source language logical forms and German logical forms. For our first machine translation experiments the source language was English. For details of the mapping process and the setup of the machine translation system see (Richardson et al. 2001a) and (Richardson et al. 2001b). A first adaptation that was necessary to make Amalgam work properly in this context was to reduce the features used in learning the models to those that are actually available on transferred logical forms. The result of retraining our Amalgam models on this smaller set of features was very encouraging: none of the models exhibited any significant drop in accuracy. The challenge of using Amalgam in machine translation is that Amalgam is trained on native German logical forms. To the extent that transferred logical forms correspond closely to native target language logical forms, the results are close to what we saw in German-to-German generation. Problems arise, however, when the transferred logical forms exhibit properties that are not found in native logical forms. Ideally, of course, the transfer component of the machine translation system should produce perfectly native-like logical forms, but this is an area of ongoing research - especially in a multi-lingual machine translation setup where the transfer component should not be fine-tuned to a specific language-pair, but should be broad and general enough to accommodate languages as different as Chinese, Japanese, German, English, French and Spanish. We are currently researching the possibilities of learning filters that post-process the transferred logical forms and adjust certain feature values to what we would expect in a native German logical form before those logical forms are input to the Amalgam module. First results on the quality of Amalgam output in machine translation compared to state-of-the-art commercially available English-to-German machine translation systems are encouraging. In April 2002, an independent agency (the Butler Hill Group) conducted a first baseline evaluation of our English-to-German translation system compared to the Saillabs system. Six evaluators compared the output of both systems (in randomized order and with anonymized source) to a reference translation. Two hundred fifty sentences from the technical domain were evaluated. The raters ranked each sentence according to a three-way distinction: 1 if the Nlpwin system was better, -1 if the comparison system was better, and 0 if both systems were equally good or bad. The result of this evaluation was that the output of both systems, while relatively poor, is rated equally: the average score was -0.069 with a +/-0.11 confidence interval at a 0.89 significance level. After implementing a prototype of the filter discussed above, adding a compound generation function and after some low-level bug-fixes, we conducted a second evaluation one month later. At this time, the average score was 0.13 with a confidence interval of +/- 0.12 at the 0.99 significance level. The number of sentences that were rated as being better in the Nlpwin system jumped from 103 in April to 131 in the second evaluation. Conclusion We have described the current state of our ongoing research into sentence realization, blending machine-learned and knowledge-engineered approaches. We are currently working with colleagues to implement Amalgam for French sentence realization. This will serve as a useful test of the extent to which the Amalgam architecture is language-independent. We continue to refine the decision tree classifiers and the ordering model. We are also experimenting with machine-learned approaches to resolving underspecified or noisy logical forms that Amalgam encounters in the context of machine translation. Finally we intend to experiment with a beam search throughout the sentence realization process, propagating the top hypotheses from each decision tree classifier rather than applying a greedy search as is presently the case. Acknowledgements Our thanks go to Max Chickering for his very generous help with the WinMine toolkit. Zhu Zhang made significant contributions to the modeling of punctuation and extraposition as an intern during the summer of 2001. Tom Reutter of the NLP group implemented the inflectional generation component for German during the German grammar checker project, and he continued to improve inflectional generation based on feedback from error analysis in Amalgam. The Butler Hill Group, and especially Karin Berghfer, have assisted in evaluation and error analysis. Last, but not least, the input from our colleagues in the NLP group has contributed much to the progress of this project. References Aikawa, T., M. Melero, L. Schwartz and A. Wu. 2001a. Multilingual sentence generation. Proceedings of the 8th European Workshop on Natural Language Generation. Toulouse, France. 57-63. Aikawa, T., Melero, M., Schwartz, L. and Wu, A. 2001b. Generation for Multilingual MT. Proceedings of the MT-Summit, Santiago de Compostela, Spain. 9-14. Bangalore, S. and O. Rambow. 2000a. Exploiting a probabilistic hierarchical model for generation. Proceedings of the 18th International Conference on Computational Linguistics (COLING 2000). Saarbrcken, Germany. 42-48. Bangalore, S. and O. Rambow. 2000b. Corpus-based lexical choice in natural language generation. Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics (ACL 2000). Hong Kong, PRC. 464-471. Bangalore, S., O. Rambow, and S. Whittaker. 2000. Evaluation metrics for generation. International Conference on Natural Language Generation (INLG 2000). Mitzpe Ramon, Israel. 1-13. Bontcheva, K. and Y. Wilks. 2001. Dealing with dependencies between content planning and surface realization in a pipeline generation architecture. Proceedings of IJCAI 2001. 1235-1240. Chickering, David Maxwell. nd. WinMine Toolkit Home Page. http://research.microsoft. com/~dmax/WinMine/Tooldoc.htm Corston-Oliver, Simon. 2000. Using Decision Trees to Select the Grammatical Relation of a Noun Phrase. Proceedings of the 1st SIGDial workshop on discourse and dialogue. Hong Kong, PRC. 66-73. Corston-Oliver, S., M. Gamon, E. Ringger, R. Moore. 2002. An overview of Amalgam: A machine-learned generation module. To appear in Proceedings of the International Natural Language Generation Conference. New York, USA. Dalianis, H. and E. Hovy. 1993. Aggregation in natural language generation. Giovanni Adorni and Michael Zock (eds), Trends in Natural Language GenerationAn Artificial Intelligence Perspective. 88-105. Duboue, P. and K. McKeown. 2001. Empirically estimating order constraints for content planning in generation. Proceedings of the 39th Annual Meeting of the Association for Computational Linguistics (ACL-2001). Toulouse, France. 172-179. Eisenberg, P. 1999. Grundriss der deutschen Grammatik. Band2: Der Satz. Metzler, Stuttgart/Weimar. Elhadad, M. 1992. Using Argumentation to Control Lexical Choice: A Functional Unification Implementation. PhD Thesis, Columbia University. Engel, U. 1996. Deutsche Grammatik. Groos, Heidelberg. Gamon, M., E. Ringger, S. Corston-Oliver, R. Moore. 2002a. Machine-learned contexts for linguistic operations in German sentence realization. To appear in Proceedings of The Fortieth Anniversary Meeting of the Association for Computational Linguistics. Pennsylvania, PA, USA. Gamon, M., E. Ringger, Z. Zhang, R. Moore, S. Corston-Oliver. 2002b. Extraposition: A case study in German sentence realization. To appear in Proceedings of the 19th International Conference in Computational Linguistics (COLING) 2002. Taipei, Taiwan.. Goldsmith, J. 2001. Unsupervised Learning of the Morphology of a Natural Language. Computational Linguistics 27:153-198. Halliday, M.A.K. 1985. An Introduction to Functional Grammar. Edward Arnold, London. Hitzeman, J., C. Mellish and J. Oberlander. 1997. Dynamic generation of museum web pages: The intelligent labelling explorer. Journal of Archives and Museum Informatics 11:107-115. Kay, M. 1979. Functional grammar. Proceedings of the Fifth Meeting of the Berkeley Linguistics Society, Berkeley, California, USA. 142-158. Jensen, K., G. E. Heidorn and S. Richardson. 1993. Natural Language Processing: The PLNLP Approach. Kluwer, Boston/Dordrecht/London. Joshi, A. 1987. The relevance of tree adjoining grammars to generation. In G. Kempen (ed.) Natural Language Generation: New Directions in Artificial Intelligence, Psychology and Linguistics. Kluwer, Dordrecht. Knight, K., I. Chander, M. Haines, V. Hatzivassiloglou, E. Hovy, M. Ida, S.K. Luk, R. Whitney and K. Yamada. 1995. Filling knowledge gaps in a broad-coverage MT system. Proceedings of the 14th IJCAI Conference, Montral, Qubec, Canada. 1390-1397 Langkilde, I. nd. Thesis proposal: Automatic sentence generation using a hybrid statistical model of lexical collocations and syntactic relations. Ms. Langkilde, I. and K. Knight. 1998a. The practical value of n-grams in generation. Proceedings of the 9th International Workshop on Natural Language Generation, Niagara-on-the-Lake, Canada. 248-255. Langkilde, I. and K. Knight. 1998b. Generation that exploits corpus-based statistical knowledge. Proceedings of the 36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics (COLING-ACL 1998). Montral, Qubec, Canada. 704-710. Malouf, R. 2000. The order of pronominal adjectives in natural language generation. Proceedings of 38th Annual Meeting of the Association for Computational Linguistics, Hong Kong, PRC. 85-92. Mann, W. and S. Thompson. 1988. Rhetorical Structure Theory: Toward a Functional Theory of Text Organization. Text 8(3). 243-281. McCawley, J. D. 1988. The Syntactic Phenomena of English. The University of Chicago Press, Chicago and London. Mel uk, I. 1988. Dependency Syntax: Theory and Practice. State University of New York Press, Albany, NY. Mellish, C., A. Knott, J. Oberlander and M. O Donnell. 1998. Experiments using stochastic search for text planning. Proceedings of the International Conference on Natural Language Generation. 97-108. Meteer, M. 1989. The SPOKESMAN natural language generation system. Report 7090, BBN Systems and Technologies, Cambridge, Massachusetts, USA. Oberlander, Jon and Chris Brew. 2000. Stochastic text generation. To appear in Philosophical Transactions of the Royal Society of London, Series A, volume 358. Penman. 1989. The Penman documentation. Technical Report. USC/ISI.. Poesio, M., R. Henschel, J. Hitzeman and R. Kibble. 1999. Statistical NP generation: A first report. Proceedings of the ESSLLI Workshop on NP Generation. Utrecht, Netherlands. Ratnaparkhi, A. 2000. Trainable methods for surface natural language generation. In Proceedings of the 6th Applied Natural Language Processing Conference and the 1st Meeting of the North American Chapter of the Association of Computational Linguistics (ANLP-NAACL 2000). Seattle, Washington, USA. 194-201. Reape, M. and Mellish C. 1999. Just what is aggregation anyway? Proceedings of the 7th European Workshop on Natural Language Generation. Toulouse, France. Reiter, E. and C. Mellish. 1992. Using classification to generate text. In Proceedings of the 30th Annual Meeting of the Association for Computational Linguistics (ACL-1992). University of Delaware, Newark. 265-272. Reiter, E., C. Mellish and J. Levine. 1992. Automatic generation of on-line documentation in the IDAS project. Proceedings of the Third Conference on Applied Natural Language Processing (ANLP-1993), Trento, Italy. 64-71. Reiter, E. 1994. Has a consensus NL generation architecture appeared, and is it psychologically plausible? Proceedings of the 7th International Workshop on Natural Language Generation. Kennebunkport, Maine. 163-170. Richardson, S., W. Dolan and L. Vanderwende. 1998. MindNet: Acquiring and structuring semantic information from text. Proceedings of the 36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics (COLING-ACL 1998). Montral, Qubec, Canada. 1098-1102. Richardson, S., W. Dolan, A. Menezes, and M. Corston-Oliver. 2001a. Overcoming the customization bottleneck using example-based MT. Proceedings, Workshop on Data-driven Machine Translation, 39th Annual Meeting and 10th Conference of the European Chapter, Association for Computational Linguistics. Toulouse, France. 9-16. Richardson, S., W. Dolan, A. Menezes, and J. Pinkham. 2001b. Achieving commercial-quality translation with example-based methods. Proceedings of the VIIIth MT Summit. Santiago de Compostela, Spain. 293-298. Ringger, E., R. Moore, M. Gamon, S. Corston-Oliver. In preparation. A tree edit distance metric for evaluation of natural language generation. Scott, D., R. Power and R. Evans. 1998. Generation as a solution to its own problem. Proceedings of the European Conference on Artificial Intelligence. 677-681. Shaw, J. 1998 Segregatory Coordination and Ellipsis in Text Generation. Proceedings of COLING-ACL. 1220-1226. Stolcke, A. 1997. Linguistic knowledge and empirical methods in speech recognition. AI Magazine 18(4):25-31. Uszkoreit, H., T. Brants, D. Duchier, B. Krenn, L. Konieczny, S. Oepen and W. Skut. 1998. Aspekte der Relativsatzextraposition im Deutschen. Claus-Report Nr.99, Sonderforschungsbereich 378, Universitt des Saarlandes, Saarbrcken, Germany. Walker, M., O. Rambow, and M. Rogati. 2001. SPoT: A trainable sentence planner. Proceedings of the North American Meeting of the Association for Computational Linguistics. Yamada, K. And K. Knight. 2001. A syntax-based statistical translation model. Proceedings of the 39th Annual Meeting of the Association for Computational Linguisitics (ACL-2001). Toulouse, France. 523-529. Zukerman, I., R. McConachy and K. Korb. 1998. Bayesian reasoning in an abductive mechanism for argument generation and analysis. AAAI98 Proceedings -- the Fifteenth National Conference on Artificial Intelligence. 833-838, Madison, Wisconsin. Wilkinson, J. 1995. Aggregation in Natural Language Generation: Another Look. Co-op work term report, Department of Computer Science, University of Waterloo.  For a subset of all possible inflected German determiners, see the variants listed in  REF _Ref531504567 \h Figure 34.  We disregard phenomena like floating quantifiers in order to keep the discussion simple.  Clearly, the downstream models could also pick up on the set of features that were predictive of the syntactic label. The use of the syntactic label as input to subsequent models, however, results in more parsimonious decision trees, e.g., for the case assignment model, as experimentation reveals.  The technical texts do not contain many examples of imperatives with exclamation marks, rhetorical questions, or other cases in which sentences would not end in a period.  Originally used to denote French verbs that form the pass compos with the verb tre.  We have also explored a different strategy for modeling extraposition: instead of successive cyclic one-step movements, it is possible to ask for each ancestor node of the parent of the cargo node whether it is a suitable target or -.=Icdmv   ; < = > ? ǾǵǬǨuluRu2jh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu hs5\jhs5U\hfhdmHnHuh&\mHnHuhImHnHuhmHnHuhUG mHnHuhfmHnHuhUG h+zhuh hsh+ hshu -.dm ) > G !!!!!!!!!!!!!!! !  ! $a$gd$a$gdsgdsgd T^jjjj? @ L M N g h i j k l m n o ҲçÔ݋q`çÔ݋ jqhxzUmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHu$jh Fhxz0J,UmHnHuh/]mHnHu j{hxzUmHnHujhxzUmHnHuhxzmHnHuh Fhxz0J,mHnHu'hxz5CJPJ\aJmHnHtH u!  # $ % & ' ( ) * + F G H I L M o p q űŦ{ӱrXDŦ'hxz6CJPJ]aJmHnHtH u2jh Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu jghxzUmHnHujhxzUmHnHuhxzmHnHu'hxz5CJPJ\aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jh Fhxz>*B*UmHnHphu   ® } rra®  jShxzUmHnHuhxzmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz6CJPJ]aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j]hxzUmHnHu      8 9 : ; < = > ? @ [ \ ] ^ a b s t u űŦ{ӱrXűŦG j?hxzUmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu jIhxzUmHnHujhxzUmHnHuhxzmHnHu'hxz6CJPJ]aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jh Fhxz>*B*UmHnHphu      ҾҰҰqҾҰWҰ2jh Fhxz>*B*UmHnHphu j5hxzUmHnHuhxzmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz6CJPJ]aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu    % & ' @ A B D E F G H I d e f g j k ³¢{aM³'hxz6CJPJ]aJmHnHtH u2jh Fhxz>*B*UmHnHphuhxzmHnHu$jh Fhxz0J,UmHnHuh/]mHnHu j+hxzUmHnHujhxzUmHnHuhxzmHnHuh Fhxz0J,:mHnHuh Fhxz0J,mHnHu'hxz5CJPJ\aJmHnHtH u    ! " $ % & ' ( ) D E ® } rra®  j hxzUmHnHuhxzmHnHu2j h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz6CJPJ]aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j! hxzUmHnHuG ' ~ BJV'H^+e0" ! " !  ! ! ! E F G H I \ ] ^ w x y { | } ~  űŦ{ӱrXDŦ'hxz6CJPJ]aJmHnHtH u2j h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j hxzUmHnHujhxzUmHnHuhxzmHnHu'hxz5CJPJ\aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j h Fhxz>*B*UmHnHphu     !";<=?@ABCD_`® } laaPl  j hxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH u2j~ h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz6CJPJ]aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j hxzUmHnHu`abghqrsŴũ~Ӵu[ŴũJ jhxzUmHnHu2jjh Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jt h Fhxz>*B*UmHnHphu()*CDEGHIJKLghijopӴӴ´uӴ[Ӵ´2jVh Fhxz>*B*UmHnHphu jhxzUmHnHuhxzmHnHu2j`h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHu" !±££uud±£ jhxzUmHnHuhxzmHnHu2jLh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu jhxzUmHnHu!"#()456OPQSTUVWXstuv{|Ŵũ~Ӵu[ŴũJ jhxzUmHnHu2j8h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu jhxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jBh Fhxz>*B*UmHnHphu !"$%&'()DEFӴӴ}rra}ӴG2j$h Fhxz>*B*UmHnHphu jhxzUmHnHuhxzmHnHu'hxz6CJPJ]aJmHnHtH u2j.h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHuFGHImnoߌr^M jhxzUmHnHu'hxz6CJPJ]aJmHnHtH u2jh Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu jhxzUmHnHujhxzUmHnHuhxzmHnHu'hxz5CJPJ\aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu&'(ABCEFGHIJefghmnʧʜ|qWF!hxzCJPJaJmHnHtH u2jh Fhxz>*B*UmHnHphuh/]mHnHu jhxzUmHnHujhxzUmHnHuhxzmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu'hxz6CJPJ]aJmHnHtH un $%׻רtc׻ר jwhxzUmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHu jhxzUmHnHujhxzUmHnHuhxzmHnHuh Fhxz0J,mHnHu %&',-<=>WXY[\]^_`{|}~Ŵũ~Ӵu[ŴũJ jchxzUmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu jmhxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jh Fhxz>*B*UmHnHphu   $%&()*+,-HIJKPQcde~ӴӴ´uӴ[Ӵ´2jh Fhxz>*B*UmHnHphu jYhxzUmHnHuhxzmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHu"~±££uud±£ jEhxzUmHnHuhxzmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu jOhxzUmHnHu CDE^_`bcdefgŴũ~Ӵu[ŴũJ j1 hxzUmHnHu2jh Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j;hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jh Fhxz>*B*UmHnHphu)*+-./012MNOPVWӴӴ´uӴ[Ӵ´2j!h Fhxz>*B*UmHnHphu j'!hxzUmHnHuhxzmHnHu2j h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHu"!"±££uud±£ j#hxzUmHnHuhxzmHnHu2j"h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j"hxzUmHnHu0`*?^' d0RY " !  ! ! ! " ! "#$*+>?@YZ[]^_`ab}~Ŵũ~Ӵu[Gũ'hxz6CJPJ]aJmHnHtH u2j$h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j $hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j#h Fhxz>*B*UmHnHphu  #$%'()*+,GH® } i^^Mi  j%hxzUmHnHuhxzmHnHu'hxz5CJPJ\aJmHnHtH u2jz%h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz6CJPJ]aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j$hxzUmHnHuHIJMNghiűŦ{ӱrXGŦ!hxzCJPJaJmHnHtH u2jf'h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j&hxzUmHnHujhxzUmHnHuhxzmHnHu'hxz6CJPJ]aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jp&h Fhxz>*B*UmHnHphu 89:<=>?@A\]±££uud±£ j(hxzUmHnHuhxzmHnHu2j\(h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j'hxzUmHnHu]^_dexyzŴũ~Ӵu[ŴũJ j*hxzUmHnHu2jH*h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j)hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2jR)h Fhxz>*B*UmHnHphu<=>WXY[\]^_`{|}~ӴӴ´uӴ[Ӵ´2j4,h Fhxz>*B*UmHnHphu j+hxzUmHnHuhxzmHnHu2j>+h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHu" !"$%&'()DE±££uud±£ j-hxzUmHnHuhxzmHnHu2j*-h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j,hxzUmHnHuEFGLMlmnŴũ~Ӵu[ŴũJ j/hxzUmHnHu2j/h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j.hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j .h Fhxz>*B*UmHnHphu   &'()/0BCD]^_abcdefӴӴ´uӴ[Ӵ´2j1h Fhxz>*B*UmHnHphu j0hxzUmHnHuhxzmHnHu2j 0h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHu")*+-./012MN±££uud±£ js2hxzUmHnHuhxzmHnHu2j1h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j}1hxzUmHnHuNOPVWdefŴũ~Ӵu[ŴũJ j_4hxzUmHnHu2j3h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu ji3hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j2h Fhxz>*B*UmHnHphu 012KLMOPQRSTopqrxyӴӴ´uӴ[Ӵ´2j5h Fhxz>*B*UmHnHphu jU5hxzUmHnHuhxzmHnHu2j4h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHujhxzUmHnHuh/]mHnHu"!"±££uud±£ jA7hxzUmHnHuhxzmHnHu2j6h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu jK6hxzUmHnHu"#$'(789RSTVWXYZ[vwxy~űŦ{ӱrXGŦ!hxzCJPJaJmHnHtH u2j8h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j78hxzUmHnHujhxzUmHnHuhxzmHnHu'hxz6CJPJ]aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j7h Fhxz>*B*UmHnHphu       " # ±££uud±£ j#:hxzUmHnHuhxzmHnHu2j9h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu!hxzCJPJaJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j-9hxzUmHnHu# $ % * + @ A B [ \ ] _ ` a b c d  Ŵũ~Ӵu[Gũ'hxz5CJPJ\aJmHnHtH u2j;h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j;hxzUmHnHujhxzUmHnHuhxzmHnHu!hxzCJPJaJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j:h Fhxz>*B*UmHnHphu b w!!"""'"t$%%[&'E).118:Z<?@;B!!!!!!! !!!! !!-gd@ >gd@ >gd@ >;gdgdn$a$gds ! " !  !!!!® } rra®  j=hxzUmHnHuhxzmHnHu2j<h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz5CJPJ\aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu j<hxzUmHnHu!!!!!"!U!V!W!p!q!r!t!u!v!w!x!y!!!!!!!!!!!!űŦ{ӱrXűŦG j>hxzUmHnHu2jv>h Fhxz>*B*UmHnHphuhxzmHnHuh/]mHnHu j=hxzUmHnHujhxzUmHnHuhxzmHnHu'hxz5CJPJ\aJmHnHtH uh Fhxz0J,mHnHu$jh Fhxz0J,UmHnHu2j=h Fhxz>*B*UmHnHphu!!!!!!!!!!!!!!!!!""""""""""##$#`#ҾҰҰqҾb^ZVPV h@ >NHh@ >hyh(jhs5CJU\aJ j?hxzUmHnHuhxzmHnHu2jl?h Fhxz>*B*UmHnHphuhxzmHnHuh Fhxz0J,mHnHu'hxz5CJPJ\aJmHnHtH u$jh Fhxz0J,UmHnHuh/]mHnHujhxzUmHnHu`#a#!$D$s$t$$$|%}%%% & &5&6&<&&&&& ' '8'9'w''''''''($(L(M(((((())D)E))'***+++++++q+r+Ͼ h)NHh/]jb@h)Ujh)Uh)h&h;F hLNH hhh&FhLh h:hiOh71 h@ >NH hH h@ >hIh@ > hNih@ >:r+++N,h,,1-8-..6.7.8.;.<.?.F.r.~... ///$/2/8///00001I2J24477Q9R999::w;x;<<<<K<Q<o=u====>>{>>>>ϾȾȾȭh7h@ >NH h7h@ >h&F h@ >NHh@ > hENH hhhEh;Fh/]j@h"XUjh&UhIh& h)NHh)@>>>>w@x@@@@@@@AAdAeA#B$B-B3B8B9BzBBBBBBBBBBBCCCCEE0F1FFFFFGGHHWIXIJJJJ!J,J2J9JJJJJKKǹhLh@ >:hIhS^Z hXh@ >hXh@ >6mHnHuh@ >6mHnHu h@ >0J< hU h@ >h71h&F hh h@ >NHh@ > h7h@ >@;B[C3D`IKN8RVJWYC[Y[ ^_`$bAbbbbc+cJcqc!!!! ! ! !! !!!!!!!!9h!v:!!!!!! & Fgd{L & Fgd{Lgdhgd~gdgdhgdugdugdS^Zgd@ >KKKKKKKLL_O`OmPnPQQQQ;S6 hI6 h@ >6 h@ >NHh&Fh71h@ >?ZZZ[C[N[Q[W[X[Y[[[O]X]a]m]]]]]]^9^=^a^^^___m_n____```````aaaaaabbbbbbƼԼ⮪jTAh!iUjh!iUh!ih[hTh[h'h- h !bNHh !b h'NHh' hiENHh5hiEh,y hhNHhhhuhnhxMhIhS^Z hS^ZNH5b b!b"b#b$b%b=b>b?b@bAbHbIb_b`babbbdbebbbbbbccddddd9e>eeeeeeeee7f8fiff³ׯ󣫯{whN4] h{LNHhhI h71NHht}h71h{LhiEh- jh~Uh~h2jAhlrh!POJQJUjw@ h2PJUVh2OJQJjh2OJQJUhh!ijh!iUh/]mHnHuh/]-qccdde$ereeiiiiii!!!!!!! !$$$Ifgdgd71 & Fgd{L & Fgd{L ffffffffffffgggHgMgkglggggggggghhhh&h'h=h>h?hFhGhHhhhhhhhhh,i4i5i6iYiciritixiiiiiiгht}h{L5h{Lh}zh h9[jh9[UjFh]:MUh. hNHh/]mHnHuh/]jFh]:MUjhUhIh71hN4]hj?jBjCjFjGjKjOjPjXjZj_j`jdjejgjhj}j~jjjjjjjjjjjjjھڮڮڮh}zhT6mHsHhThT6 h}zhThTh}zhT6ht}ht}6h9[h71h{Lh}zh{L6mHsHh}zht}6mHsH h}zht}h}zh{L6ht} hDKg5ht}ht}5 hi; 53iiijj jjjjj /$IfgdjkdG$$Ifs4%"" t0644 saf4 j j4jHY<! $$$IfgdkdH$$Ifs4rxc N9%" t0644 saf44j5j:j?jCjGjPjXj`jejfj /$IfgdjkdH$$Ifs4%"" t0644 saf4 fjgjjHY<! $$$IfgdkdI$$Ifs4rxc N9%" t0644 saf4jjjjjjjjjjj /$IfgdjkdI$$Ifs4%"" t0644 saf4 jjjjjjjjjjjjjjjjjjjjjj k k k'k(k*k.k/k3k4k7k9kyCyDyLyMyqyryyyzzzzzzzzzzzzz{{ { {{{x|||||||ӽyynhEh7 mHsH hh&hEhhGh !bmHsHhGh7 mHsH hUhkEh !bhU hUhUhUh7 NH hUh7 h/]jSh]:MUh7 jh7 Uh[v-h71 hkENHhxMhkEjh[v-0JU+ {i{{|x|||3}O}~}}nɀ+i#!!!!!!!!!!!!!!!!!!!!!!!gdT 80^8`0gdn$gdlgd !bgd !bgd7 gd 80^8`0gdE|||||}}1}2}3}A}B}M}N}g}h}|}}}~}}}}}}}}}}}} ~~~XYv|./mnƿƿƿͻͷͳ} h*5hT h !bNHh&Fhvh7 h: h NHh hEhs hUh !bhEh hEh7 hEhE hd:h7 hGh7 mHsHhGhEmHsHhEh7 mHsHhEhEmHsH0ȀɀЀр*+34ghipqŁƁȁ́!"#$:;<CDĽ鶫̫̫Ķ鶙wsjh/]mHnHuh/]j\Th 2Uh 2jh 2UhEhNmHsHhEhp6emHsH hEhp6ehGh*mHsH hEh* h:|h*h:|h*5hGhNmHsHhGh mHsH hEh hEhN hEh !bh*h*5)DÊ͂ӂz{ĆΆFMNdefgijw&:;ˆȈɈˈ̈ψЈֈ :ܳܯ󡚡ꊡh{.hiOhO ,hYh7 hhh.;Y hvNHhxMh{hNh/]mHnHujhp6eUhp6ehp6ehp6eNH hp6ehp6eh]:M h 2NHhv h:hEh 2jh 2U6{}І! $$IfgdygdT ]wPPPP $$IfgdykdT$$If44\iOsss07644 laf4%+^\ QQQQ $$IfgdykdPU$$If4\iOsss07644 laf4+,39?E^/SSSS $IfgdykdU$$If4\iOsss07644 laf4EF̈ވ a^/Y!v:T!O!J!E!@!gdgdVEmgdO ,gdgd{gdp6ekd8V$$If4\iOsss07644 laf4$&'-.23<=֊78NOPWXYhr =>TUV]^_`abcдУИДj#XhVEmUhjWhUj)WhU h&mNHh&mh/]mHnHuh/]jVhUhJ]jhJ]Uhah71 h:hiOhVEm hSVhVEmh.;Yh{.7ac1H_yO˕d!l!v:!>vA!v:!!!,!v:!!!!!!!!!!$ & Fx$Ifgd$pgdbgdugdugdSV^gd & F@gdgd gdO ,gdJ]gd $$a$gd!icjkÌČڌی܌݌ߌ$1JPzɍҍ 01ghopŽ܎ڼĞĞhIh71 h_^NHh_^hF! hbhF! hbhb hbhd: h 6hd:huhO , hJ]:j1ehJ]U h: h:hiOhVEmhah/]mHnHujhJ]UhJ]78<BCRΏϏЏ܏ݏ !"#%&56=>GHVejÐǐː̐АѐՐ֐ﬨ礞뤑jheU he6 heNHhehhhhajzh Uh/]mHnHuh/]jEzhRUjh U h h hd:h71h hI:hIh_^49>?Cbcijm\]cdky_`LMOϔ2:ʕ˕"ijį~ h:hT: hjMhuhshB' h{.NHhSVh{.h6hhAohNhO ,h h:hiOh h71 he6hI heNHheh/]mHnHuh/]jheUj]heU0"?@AVWX_`abcdїES'(,-Ii5<=STUVWy'(/<ƾhnXd hbhbhb h]NHh] hhjhe%60JU he%6h6 he%6NHhe%6h$rhI h:hbhT: h/]mHnHuh/]jڎh$rUjhbUhb h:he%63 Da7aݘGz™Й/j{!!!!!!!!!!!!!!!!!!!!$ & F $Ifgd$r$ & F $Ifgd$r$ & F $Ifgd$r$ & F$Ifgd$r$ & F$Ifgd$r{Ț45y!!!!fa\ !W!v:R!gdTgdygde%6gdbkkdU$$IfTl," t0644 laT & F$Ifgd$r$ & F$Ifgd$r$ & F $Ifgd$r %qoqū'ѯ:"[!!!!!v:! !!!3!v:!!!!!!!!!!!! ! & Fgd%gd%gd2+gd Rgd RgdTgdz $$a$gd!igdgdTʠˠ̠ѠҠ2EUYjosޡߡ"#67YtϢТѢҢӢȼ᦮jhzUjhzUjhOUhO h2+NHhphbhbhp6hbhT6 hTNHjhzUh/]jhzUjh2+Uh2+hT8Ӣ %8DJirs NYirw~ݤޤ8Ĺ⬣h RmHsHjhOUhOhOmHsHjh RU h6h h36h3h36h/]mHnHsHuhOh/]mHsHjh,wUjh3Uh3h RhT hONHhbhIhO.8pqեޥfqʦԦ&'-.>?ŧا٧ gpvרبǿǴޣ␛ڈ~h3h36NHh3h36jh.lUjh RUhh/]mHnHuh/]jh.lUjh%Uh% h RNH h:hah3hIh RhThz6mHsHhbhO6mHsH1$,.`ev٩ 678?@AŪ̪ nopxyūЫի'INklWXnoǿǷǷϷdz˯˯˥jdh.lUjh Uh h%NHh%h2+jh RUj h RUh Rhkh/]mHnHuh/]jh,wUjh3UhIh3h3h369opwyzʭ %-9;<DEJOPT`bchinop®î%&'(Zuͯϯh%Th%6huhiOhT h%NHh71hwhkh%Th%h h/]mHnHuh/]jh UHϯЯѯүگ߯-2689:;LW~ǰȰΰհް߰ !">ֱױ12@GINOmDzȲвѲhiOh%Thw6 hwNHh hhu h%NHhkh71hwh&Fh%Th%6h%h%TJ[\de{|~дӴشٴqrʵϵٵLUpq~ ݦjh.lUjh:0Uh:0hcI h NHh2+h%h&Fh%Tj^hwUh/]mHnHuh/]jh.lUjhwUhwh hk=[]egɼmOQa!!v:!!!&!v:! !!!!!v:!!!gdT>gd/\gd&gd:0FEƀ[fFgdwgd2+gdw $$a$gd!i678=>e{ƹǹܹ %&<=>EGHĺ!"()efnoɼмѼ׼ؼ೾䪾ঢྪئh2+hajbh:0Uh71h&h/]mHnHujh.lUjh:0UhcIh:06hI hcINHhcIhTh:0h/]jjhRp1UhkjhkU9ؼ[\abԽ1278wx}Ŀ klm(+,-FGOouɿɿ޿޿jh/\U h/\NHh/\ hy7NHhy7 hh hNHh&FhhcIhkh h7k9NHh7k9h71h&A *+7>NOPXYoprsuv #14>hpqr Trs۽۽۹jhT>Uhnk7hGhiOhB'hJFh#OhT>h2+h&Fhjh/\Uhy7hk h/\NHh/\h/]mHnHuh/]jh/\Ujh.lU7 BCnsw+137`a~VtG <=ƾƳƭƩХh|lh hvxNHjh:UjhvxUhvx hNHhGhJFhB'h71hhT>h/]jhT>Uj%h.lU>=>STU\^_`abjkFijpq ׿׻󷦷ېjh.lUhv1jhv1U hvNHh71hv hDgNHjhhUhDgh2+hjhGUjhGUhT>hGh/]mHnHuh/]jhUhgjhgU2ac[-!!v:! !!J!v:!!!!9h!v:!!!$!v:!!c!v:!!$gdgdgz. $$a$gd!igdgz.gd{Vgdm2ygdv1gdv1gdGgd2+gdG $$a$gdYh7 679:<=Z[ (*Cop'57UW`z{ǿǥә h"XNHh"X hN<_NHhN<_h,whN<_6 hN<_hN<_h,w h,wh,wh,wh,w6h,whv16 h,whv1 hm#U6hm#Uhvhh/]mHnHujhv1Uj`hv1UhGhv14S+,HIMNXYklabghi./ !/[\rst{}~߽ѹ񑹱jKh{VUh/]mHnHuh/]jh.lUjh{VUh{VhiOjh,w0JU hcNHhch71 hONHhOh,w hm2yNHhJFhm2yh2+ hN<_hv17.<+,-467bf}~567>@AúúҺújh.lUh,wjb hgz.Uh71j h.lUh/]mHnHuh/]jh h.lUjhgz.UhJFhh5 hgz.NHh&Fhgz.h2+h{Vh l:,-lm*+,-./GHIKcdzi!jo@ hYCJPJUVaJjQhgU!jo@ hgCJPJUVaJhEj&hjU!j`? hjCJPJUVaJhP<jhP<UhjM h6h&FhTh,wh/]mHnHujhgz.Ujhgz.Uh,whgz.6NHh,whgz.6hgz.).g<8<`uK!!!!!!!!!!!+!&)T!!!!gdM gdM gd $$a$gd6+5gd & Fgd:|gdagdOgd2+$gdO $$a$gddefgno$@K %+:Tx23V\ EFLM[ҾҾҸҾҴҰҸҬҸҬҨҘ h6+5hjhUhamHnHu hhhbhhIh&F haNHhJFhiOh:|hgha6hahOh!ih/]mHnHuhP<hEjhP<Uj7hYU4)LMNdefmop<_`Kžűߥ챊챆~x~ h"NHh"hOhM jghUjhUjhTUjhTUhThhhNH hhh6+5 hVOhhVOh6 h6+5hh/]mHnHuh/]jhUjhU/Kgiy&F?DgvQ  ! !!!!!!!!!!!!!! !!!!gd9gd9gdgdgd8Kgd8Kgd_gd_gd+gd+gdM gdM *+:;I~$>O5hij&'wx @h_h'h&Fh" h+NHh+h71hM hM NHS %&]^mnAayz}~󺴺󺧺hThVOh71h&F h8K6 h8KNHh8KhJFh+hyQ h_NHjh:5{Uh/]mHnHuh/]jh:5{Uh_jh_U>Sbv{ `diotBCD  &'=>?FHIY,7YaǾǾ亢 h9NH h96j hRp1UjhJFUh9h/]mHnHuh/]jh:5{UjhU hNHh&Fh&Chhk+hJFhh8K=19A[} ,-34PQ 4 F I n q       1 2   p    !*<=BCDEKL`ag|h71h&Fh" hh/gNHhh/ghg} hptNH hk}6hk} h)!NHhpt hNHhyQh)!hheih9hVOE    _D\$9! !!!!!!!!!!!!! !!!!!!^gdW9gd&C:gd&CgdLXUgd2+gdagd2+gdh/ggdOgd"gd"gdg}gdg}gdptgdptPQ ?@  OVWZ]^^_[dCD\  pqw?ͽ hahahk+hRp1huh2+ hh/ghh/g h:hiOh'ehO hg}hoho h"NHhP!hVO hh/gNHheihh/gh"Ab  XY`r~HIbEFL^_#$%79:ÿǸôêæêÿǟǿ hWNH hWhWhVO h=NHhhiO hWh=hWh=hY?hXhyhbsh&C hLXUNHh&Fh#whC hm$WhChP!hLXUh2+<         !!!/!:!Z!^!!!! " """"$"("*"/"1"6"J"N"Z"_"j"q""""""""""""""" ##)#1#7#8#E#G#N#`#c#w#z#|###콵hVOhhchF\5hF\hhch"L5hEh"L5h! h"LNH hl6hlhl6hlh"Lh#whY?h&ChWD    ! ""7"J"#$Q%i%?(S(((/+F,k,!!!!!!!!!!!!!! !!!! !!! & Fgd)gdW^gd:5{^gdE^gdhc^gd"Lgd!9gd&C^gd&C:gd&C##############$ $ $$$$.$/$2$3$6$9$;$?$A$E$G$J$O$P$W$\$k$o$$$$$$$% %(%*%,%2%@%C%N%P%Q%i%k%|%~%%%%%%%&&*&,&B&D&k&m&&&&&&&&' hhcNHhhchF\5h"Ljh+40JUh71hhchhc5hhchF\ hF\NHP'9';'t'v'''''''((?(S(V(b(n(((((((((((() ))) )!)+)B)C)K)S)T)\)a)b)c)g)t)y))))))))))))))))))**'*+*8*@*B*G*U*X*q*u* hEhE hENHhiOhEhE5h&C hE5hEhhEhhc5hhchhc5hhcLu**********************++!+.+/+J+r+~+++=,>,D,E,F,Q,S,~,,,,q-r-|-}------z...//111+1-1.191:111d2¾ҾҾh&FhLXU h-hKxhP!hKxh0U,h'Hh-h2h#w h)6 h)NHh)hW hEhEh:5{h:5{hE5hECk,---z./:1M1X1225333344I4757B9!!!!!!!!!!!!!!!!! !!!gdT$a$gd agdB9 & F!gdh%h & F!gdh%hgdLXUgdLXUgdLXUgd:5{ & Fgd)d2e2222435363X3Y3333333344|4}44444444444555555<5=5K5L5S5T5j5k5555555555555555566&6'63646<6=6J6K6Q6R6o6p6y6z666666666hiO hB9NHhB9h3phyoWhl hLXUhLXUhLXU hLXUNHT6666677%74757W777'8+8M8i8l8r888 9 9 99999999#9$9-909@9A9E9W9X9G:*;========>&>:>;>Q>ʻҙʲʕh6+5hThTmH sH hR hTNHhkiDh/]mHnHuh/]jh'`UjhTU hThThA(h#whbshyhThLXU hB9hB9hiOhB96B9F9P9W9a9b9g9w99''L''kd$$IfTl\765% t0644 laT$$$Ifgd?s9999999L''kd$$IfTl\765% t0644 laT$$$Ifgd?s999999ZLL'L'L$$$Ifgd?skd$$IfTl\765% t0644 laT999:::ZLL'L'L$$$Ifgd?skdr$$IfTl\765% t0644 laT::%:1:?:F:ZLL'L'L$$$Ifgd?skd$$IfTl\765% t0644 laTF:G:L:\:l:s:ZLL'L'L$$$Ifgd?skdd$$IfTl\765% t0644 laTs:t:x::::ZLL'L'L$$$Ifgd?skd$$IfTl\765% t0644 laT::::::ZLL'L'L$$$Ifgd?skdV$$IfTl\765% t0644 laT:::::;ZLL'L'L$$$Ifgd?skd$$IfTl\765% t0644 laT;; ;;";);ZLL'L'L$$$Ifgd?skdH$$IfTl\765% t0644 laT);*;/;?;O;V;ZLL'L'L$$$Ifgd?skd$$IfTl\765% t0644 laTV;W;\;l;|;;ZLL'L'L$$$Ifgd?skd:$$IfTl\765% t0644 laT;;;;;;ZLL'L'L$$$Ifgd?skd$$IfTl\765% t0644 laT;;;;;;ZLL'L'L$$$Ifgd?skd,$$IfTl\765% t0644 laT;;;;< <ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laT < <<#<7<><ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laT><?<D<X<l<s<ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laTs<t<|<<<<ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laT<<<<<<ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laT<<<<<<ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laT<<< ==&=ZMM'M'M $$Ifgd?skd{$$IfTl\765% t0644 laT&='=-===M=T=ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laTT=U=Z=n===ZMM'M'M $$Ifgd?skdm$$IfTl\765% t0644 laT======ZMM'M'M $$Ifgd?skd$$IfTl\765% t0644 laT==>&>???n@ZU!v:P!K!F!P!A!gdkiDgdLXUgdTgdLXUgdTkd_$$IfTl\765% t0644 laTQ>R>S>Z>\>]>>>>>>>???????m@n@}@@@ A AKA]AeAA'B:BIBJBbBdBeBkBBBBBBBBBBBBľ۷z߅jUh'`UjhkiDU h3_h3_hA(hi>shyhyoWh3_ hkiDhkiDhkiDhLXU hThT hRNHhR hTNHh h:hiOhTh/]mHnHuh/]jhTUjh'`U0n@}@@@ AKA]AA'BJB2C6C@CGCQC!!!!!!!!!!.cc $$IfgdN8$a$gd agdkiD & FgdkiD & Fgdh%hgdLXUBBBBB C1C2CRCXCfCsC{CCCCCCCCCCCCCCCCC DDDD'D4D@DGDSDTDUD\D]DsDtDvDwDDDDDxEyE$F%FDFFFFFFkGlGyG쾺hyoW hbshbs hRhR hbsNHhbshLXUh6+5h/]mHnHujhkiDU hkiDh3_ h?O h3_ hkiDhkiDhkiDh4whRhi>sh3_>QCRCXCmCCCZM.McMcM $$IfgdN8kd$$IfTl\  t0644 laTCCCCCCZM.McMcM $$IfgdN8kdC$$IfTl\  t0644 laTCCCCCCZM.McMcM $$IfgdN8kd$$IfTl\  t0644 laTCCCDDDZM.McMcM $$IfgdN8kd%$$IfTl\  t0644 laTDD'D:DMDTDZL.LcLcL$$$IfgdN8kd$$IfTl\  t0644 laTTDUDDD%F9FDFFZT!v:O!J!E!O!@!gdbsgdLXUgdRgdLXU$gd3_kd$$IfTl\  t0644 laTFFF7GkGGGGGHHII&I-I7I!!!!!!!!!!!'' $$Ifgd l$a$gd agd & FgdbsgdLXUyGGGGGGGGHHHHHHHHHHHI9IIJ5J7I8IEIUIeIlIZMM'M'M $$Ifgd lkdx$$IfTl\edc0 t0644 laTlImIrIIIIZMM'M'M $$Ifgd lkd$$IfTl\edc0 t0644 laTIIIIIIZMM'M'M $$Ifgd lkdZ$$IfTl\edc0 t0644 laTIIIIIIZMM'M'M $$Ifgd lkd$$IfTl\edc0 t0644 laTIJJJ-J4JZMM'M'M $$Ifgd lkd<$$IfTl\edc0 t0644 laT4J5JJJKKKLZXS!N!I!D!?!gdgOgd~gd~gdLXUgdLXUkd$$IfTl\edc0 t0644 laTLLM?MfMMM NO1O PPP!P+P!!!!!!!!!!_'' $$Ifgd?s$a$gd agdgO & FgdgOgd~LLLMMfMgMMMN OO1O2O:O=OVO\OOOO P PPPPPPPPPP#Q$Q5QQQQQQQQQQQARBRjR S S7SPTQT`TTTUU(U:UCU V VżŸh4wh a h|h|h|hEh71h6+5h/]mHnHujh,KU hgOh,K h,Kh,K h,KNHh,KhhA( hgOhgOhgOhyoWh~>+P,P1PAPQPXPZO_O'O'O $IfgdyoWkd$$IfTl\  t0644 laTXPYP^PrPPPZO_O'O'O $IfgdyoWkd$$IfTl\  t0644 laTPPPPPPZO_O'O'B $$IfgdyoW $IfgdyoWkd$$IfTl\  t0644 laTPP$Q5Q S,S7SQTZUP!K!F!P!A!gd|gd~gd,Kgd~gd,Kkdq$$IfTl\  t0644 laTQT`TTTU(U:UhU V.VVVVW W!!!!!!!!!! $$Ifgd?s$a$gd agd| & Fgd|gd~ V.VMVPVXViVoVqVVVVlWsWtWWWWWWWWW0XIXXXY[Y^YYY;Zբ&ä#Fåͥԥޥ!!!!!!!! $$IfgdyoWgdgdv9gd~gdKx & F-gd ޥߥ ZOOOO $IfgdyoWkd $$IfTl\  t0644 laT*<CZOOOB $$IfgdyoW $IfgdyoWkdD $$IfTl\  t0644 laT$%1267DKLbcefqݦͧ +,EFGJKvwjkީߩԪժתتfؿԜh71 h|NHh|hXh/]j& h*UjhO(Uhf/hO( hv9hv9 h\NHhkh#wh~h6+5h/]mHnHujhUh hhh\9CD§̫ͧ۫ZXS!N!I!S!NS!gd~gdv9gd~kd $$IfTl\  t0644 laTfgijɫʫ˫̫۫*.2589AMN"\]qޭ߭RSghiԮծ+.xyүӯ/ABXۼۼۼ۸h;O hO(hO(h$/h ahv9h&FhFhf/hO(h~ hv9hv9jhZ0JUhZh|h71H۫Nrh +ӯBְ-?KV!!!!!!!!!!!!!!!!!!!!!!gdve$a$gd6+5gdv9gd~ & F.gdO( & F.gdF & F.gdO(ŰǰȰʰ˰հְװ!",-?D}ӱԱIJK-1ADUVYZ[deklֹıĪ hKxh;O hKxhKxhN8hZ hC\MhC\MhFhKxhA(hv9 h$/NHh$/he%6h~ hv9hO( hO(hO(h71h;OhO(hf/@V[elvwzssOsskd $$IfTl\G _ t0644 laT $IfgdyoW luwyzŸ̸ѸҸظٸڸ.?׹ع./67>EWk `ƽƹεh* hv9hv9h;Oht3h71 h|NHh|h~h6+5h/]mHnHujhKxUhKx h%h$/ hKxh$/h$/ huh$/ h h$/ hKxhKx hKxh;O:ҸٸZOOsOsO $IfgdyoWkd $$IfTl\G _ t0644 laTٸڸ.? ѼZXS!N!I!S!N!S!gd~gdv9gd~kd $$IfTl\G _ t0644 laT`aмѼ&'9:Z[m  0=?@GZ`¿ÿտֿؿٿ"#1245?@BCNO`acdnoqr~h*mHsHh*h*mHsHhW h*h*hA(hyoW hv9h*h ahv9h~ hv9hv9h* h*NHH&9[m 0!!!!!!!!!! $Ifgd a$a$gd6+5gdv9gd~ & F/gd*¿οܿYNNNN $Ifgd akd $$IfTl4\( t0644 laT !ZOOOO $Ifgd akdj $$IfTl\( t0644 laT!"*8FMZOOOO $Ifgd akd $$IfTl\( t0644 laTMNYgv}ZOOOO $Ifgd akdL$$IfTl\( t0644 laT}~ZOOOO $Ifgd akd$$IfTl\( t0644 laT!"-.12>?OPRS]^`almxy~!"/056DERSXYfgh*mHsHh*h*mHsH^ZOOOO $Ifgd akd.$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laT&6=ZOOOO $Ifgd akd$$IfTl\( t0644 laT=>HVdkZOOOO $Ifgd akd$$IfTl\( t0644 laTklqZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akdc$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laT(<CZOOOO $Ifgd akdE$$IfTl\( t0644 laTCDK_szZOOOO $Ifgd akd$$IfTl\( t0644 laTglm{|   !1245?@BCOP]^cdqrwx  *+/0<=ABOPh*mHsHh*h*mHsH^z{ZOOOO $Ifgd akd'$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akd $$IfTl\( t0644 laT *8GNZOOOO $Ifgd akdz$$IfTl\( t0644 laTNOVj~ZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akd\$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akd>$$IfTl\( t0644 laT#5GNZOOOO $Ifgd akd$$IfTl\( t0644 laTNOZhw~ZOOOO $Ifgd akd $$IfTl\( t0644 laTPabdeoprs  ,-12?@NOST`aefst h* h*h*h*mHsHh*h*mHsHY~ZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laT ZOOOO $Ifgd akds$$IfTl\( t0644 laT  %7>ZOOOO $Ifgd akd$$IfTl\( t0644 laT>?GYkrZOOOO $Ifgd akdU$$IfTl\( t0644 laTrszZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akd7$$IfTl\( t0644 laT ZOOOO $Ifgd akd$$IfTl\( t0644 laT  !07ZOOOO $Ifgd akd$$IfTl\( t0644 laT  ()+,89JKMNXY[\ghstxy'(569:FGJKXYhilmxy|}h*h*mHsHh* h*h*h*mHsHY78CQ_fZOOOO $Ifgd akd$$IfTl\( t0644 laTfgl~ZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akdl$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laT &ZOOOO $Ifgd akdN$$IfTl\( t0644 laT&'.?PWZOOOO $Ifgd akd$$IfTl\( t0644 laTWXaqZOOOO $Ifgd akd0$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laTZOOOO $Ifgd akd$$IfTl\( t0644 laT "#&'2367CDSTWXcdghtu#$)*89RSUV`acdoph*mHsHh*h*mHsH^ ZOOOO $Ifgd akd$$IfTl\( t0644 laT+;BZOOOO $Ifgd akd$$IfTl\( t0644 laTBCL\lsZOOOO $Ifgd akde $$IfTl\( t0644 laTstzZOOOO $Ifgd akd $$IfTl\( t0644 laTZOOOO $Ifgd akdG!$$IfTl\( t0644 laTZOOOO $Ifgd akd!$$IfTl\( t0644 laT07ZOOOO $Ifgd akd)"$$IfTl\( t0644 laT78KYgnZOOOO $Ifgd akd"$$IfTl\( t0644 laTno{ZOOOO $Ifgd akd #$$IfTl\( t0644 laT  )*,-89FGJKVWZ[gnoڶhJ^ h*NHh/]j%h*Uh~hN8h6+5h/]mHnHujh*Uh* h*h*h*mHsHh*h*mHsHDZOOOO $Ifgd akd|#$$IfTl\( t0644 laT ZOOOO $Ifgd akd#$$IfTl\( t0644 laT  "07ZOOOO $Ifgd akd^$$$IfTl\( t0644 laT78?O_fZOOOB $$Ifgd a $Ifgd akd$$$IfTl\( t0644 laTfgZXS!NI!S!N!gd~gdv9gd~kd@%$$IfTl\( t0644 laT+,89C,2qr   %&)*5ۺhmHsHhhmHsHhH` hhhA( hv9hhh ahv9 hJ^NHhJ^he%6h~ hv9hv9F8rv!!!!!!!!!!!! $Ifgd a $$Ifgd.b$a$gd6+5gdv9gd~ $ & F0gd.b $$ & F0gd.b$gd.bYLAAA $Ifgd a $$Ifgd.bkd.&$$IfTl4\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkd&$$IfTl\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkd'$$IfTl\w  t0644 laT.>EZMBBB $Ifgd a $$Ifgd.bkd'$$IfTl\w  t0644 laT569:FGUVXYcdfgst~ ()459:FGKLYZefklyzhhmHsHhmHsH^EFN\krZMBBB $Ifgd a $$Ifgd.bkd'$$IfTl\w  t0644 laTrswZMBBB $Ifgd a $$Ifgd.bkdf($$IfTl\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkd($$IfTl\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkdH)$$IfTl\w  t0644 laT 'ZMBBB $Ifgd a $$Ifgd.bkd)$$IfTl\w  t0644 laT'(-?QXZMBBB $Ifgd a $$Ifgd.bkd**$$IfTl\w  t0644 laTXY^rZMBBB $Ifgd a $$Ifgd.bkd*$$IfTl\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkd +$$IfTl\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkd}+$$IfTl\w  t0644 laTZMBBB $Ifgd a $$Ifgd.bkd+$$IfTl\w  t0644 laT  ,-/0:;=>IJYZ\]ghjkvw  ()89;<FGIJUVcdhiuvz{h hhhhmHsHhmHsHY%3AHZMBBB $Ifgd a $$Ifgd.bkd_,$$IfTl\w  t0644 laTHIR`nuZMBBB $Ifgd a $$Ifgd.bkd,$$IfTl\w  t0644 laTuvZOOOO $Ifgd akdA-$$IfTl\w  t0644 laTZOOOO $Ifgd akd-$$IfTl\w  t0644 laTZOOOO $Ifgd akd#.$$IfTl\w  t0644 laT 'ZOOOO $Ifgd akd.$$IfTl\w  t0644 laT'(1?MTZOOOO $Ifgd akd/$$IfTl\w  t0644 laTTU\nZOOOO $Ifgd akdv/$$IfTl\w  t0644 laTZOOOB $$Ifgd a $Ifgd akd/$$IfTl\w  t0644 laT{#4?@FG\hOPefkl~klǿǻǵDZǥǡǡhP-h(hXEheWL6heWL hwdNHhwh'h71hwdh, hv9hv9 hNHh~h6+5h/]mHnHujhUh hh?#4G]hZXS!N!I!S!D!gdeWLgd~gdv9gd~kdX0$$IfTl\w  t0644 laTY4J$hxY]gnx!!!!!!!!!!!!!!!ss $$Ifgd.b$a$gd6+5 & F8gd( & F8gdwdgd~gdv9%XY347JS$+ghx!#(.46:XYyz»»» h(hP- h+9BhP-h[ h(6h' hP-hP-hA(he%6h~ hv9h( h(h(h ahP-h(hv9Gxy~ZMMsMsM $$Ifgd.bkd0$$IfTl\"m  t0644 laTZMMsMsM $$Ifgd.bkd:1$$IfTl\"m  t0644 laTZOOsOsO $Ifgd akd1$$IfTl\"m  t0644 laToQ2>IGZXS!N!N!I!S!N!gd~gdv9gd~kd2$$IfTl\"m  t0644 laT%&()4Shno1PQ12IFG 67QR{|1>̻зȷС hv9hMEhNTh1hR$h71 hMENHhyhME hv9h$a h$aNHhv9h$ah~hnh,h'h6+5h/]mHnHujh(Uh( h(hP-9GE5xC?QZ}!!!!!!!!!!!!!!!!!!!!!!  $$Ifgd & F:gd1gdv9gd~gdME & F9gdME%Be>?QBDEOPQYZ}  "#23ABDEOPRS_`lmopz{}~ɼԼ h3h3hD,hA(hTg hhv9he%6h3hhNH hhhh~hv9h1mHnHuhmHnHuhNTmHnHuh1mHnHu>*1 M  kd2$$IfTl\! t0644 laT $$Ifgd12:HW^ZMM M M $$Ifgdkd2$$IfTl\! t0644 laT^_esZMM M M $$Ifgdkdo3$$IfTl\! t0644 laTZMM M M $$Ifgdkd3$$IfTl\! t0644 laT  +23IJLMX "#1289HIUVXYcdfgsthD,hiOhTghnh6+5h/]mHnHujh3Uh3 h3h3RZMM M M $$IfgdkdQ4$$IfTl\! t0644 laT#*ZMM M M $$Ifgdkd4$$IfTl\! t0644 laT*+ ZXS!FF F F $$Ifgd.bgd3kd35$$IfTl\! t0644 laT  *@GZOO O O $IfgdyoWkd5$$IfTl\! t0644 laTGHN\krZOO O O $IfgdyoWkd6$$IfTl\! t0644 laTrs~ZOO O O $IfgdyoWkd6$$IfTl\! t0644 laTZOO O O $IfgdyoWkd6$$IfTl\! t0644 laT ZOO O B $$IfgdyoW $IfgdyoWkdh7$$IfTl\! t0644 laT|yaZX!v:S!N!I!D!? !gd&gd&gd2+gd~gd~kd7$$IfTl\! t0644 laT2356A{|xy opqЕ hUhUhUjD9hUjh"6UhiO h"6NHh"6h& h( h&h2+ hLXUhLXUh Zj8hk$Uh/]jJ8hk$Uh~hnh6+5h/]mHnHujh3Uh31 px|}IJ0L`abjkʮ h[NpNHh6+5h6+56h6+5hYv6 hrlhYvjh[NpUh[Npj>:h[NpUhiOhnk7h/]mHnHuh/]j9h7]Ujh7]Uh7]hUhh& h&NH6acEq'=!!v:!!!!!!H!H!!H!H!H!H!!H!H!!H & FCgd6[gd6[ & FBgd6[gdv & F4gd& & F4gd_ & F5gd&gd&gd&gdgd[Np$a$gd6+5  ;<<Xkl $OPQ )34VW(DK_ÿ߷׿׿ӫӫӫhvh_h6[h71 h&NHhwh&hnhYvjh[NpUh[NpjRHh[NpUhvehnk7hiOhh7]h/]mHnHuh/]jh7]UjGh7]U9_qUVXYJKnot  #$WXYdq!"<=wȷԫhnh/]mHnHujhUhj Vh "U!jm2? h "CJPJUVaJhlJ$jhlJ$UhYvh7] h&NHhw h6[NHhiOh6[hvh_h&:== <lw -    M!!=(!v:!!e|!1!1!1!e|!v:!!!!!!!!gdGgdugd!>gd{Vgd&gdgd6[&$d%d&d'dNOPQgdgd$a$gdmagd&56no ץˀˀۀjh6[Uhh&OJQJ^JhhhOJQJ^JhhOJQJ^Jh\2hjhh7]U hNHh6[hiOh&hh/]mHnHuh/]jhh7]Uh7]jh7]U0vw   8 9 = > M N S T \ ] v w x y        . /   W X d e           Q W c d t ~     *45Vÿø h7hGhGhu h!>h!>h?sh1>h{VhFNhPy h&NHh& h( h&hh71h7]hiOhhnhwBV\kwz  ()glno&'12!+13ABCMno)*^qx|}[\jhGUh|hCEhTg h]hG hGNHh7hGNHjh+d0JU h7hGhwhGHMja{9 "%v''(()*!!!!m1!s!s!m1!s!!! ! !!!!!!!gdigdugdSMgd)4gdG83$$$ 0$d%d&d'dNOPQ^`0a$gdGgdG :;jnz|}XY`abeor]^rszعhhGNHnHtHhGCJnHtHhhGnHtHhGnHtH h: hG hGNHjhG0JUhGh/]mHnHu h7h/]jhGUj|ihGU7z{FG1679Qpqr   N!O!!!! """"IJIJIJIJhx8hUvh3BhXphCE h~pNHh~p hSMNHh[hSMhVOh)hZ]hGhDRh/]mHnHujh7hGU h7hGhhGCJnHtH<"""""###@#H#h#s####$8$:$Z$$$$$S%T%e%%%%%&&n&}&&&&&&&.'4'u'v'w''*))))) ****** +y+++,-޵ֵ޵ֵh{Kh?sh[fh.e hihk+hk+huhZ] hSMh~phd!h hUvNHhCEhhihX3h%ThUv h~pNHh~phRp1h3v>**n-y-2../0;11h2)344566678k99u:;;X<! !!!!!!!!!!!!!!!!!!!!!!!!gd@ >gdYh7gdP!gdGgd@ >;gdgd.egdu--m-n-y-----...2.i............./C/E/r/s//000L0M0j00000 1 11Ľཱུ౭ज़ज़|p|hGhG6NH]hGhG6]hTg h7hGh {h@ >6NHh {h@ >6H*h {h@ >6hkEhR7h19h@ >6 h/ih@ > hU h@ > h@ >NHhD1h@ >6H* h@ >6h@ > hh@ >hu h.eh.eh[f h[fNH+11;1111122.2/2I2J2g2h2222 3 333"3(3)3\3c3334%4P4Q4y4ټvncnXhR7hR7nH tH h,h@ >nH tH h@ >nH tH hhYh76]hYh7 hhYh7h@ >nH tH h+4h@ >6NH]nH tH h@ >6]nH tH h+4h@ >6]nH tH h+4h@ >nH tH hkE hOhP! hfhP!hP! h(!KhP!h {h@ >6h@ > h7hGh7hG]!y44444444=5R5T555555555566616q6r6666666667ū}ukuū_X hhYh7h_^h@ >6mHsHh {h@ >6NHh {h@ >6 h?sh@ >h {h@ >6mHsHh-2h@ >mHsHh@ >mHsHhR7mHsHh_^h@ >mHsHh~|h@ >6H* h@ >6h@ >h+4h@ >nH tH hR7hR7nH tH hkEnH tH hkEhkE6nH tH hkEhR76nH tH "77|77777888888E9_9j9k99999999*:3:>:h:t:::4;c;;;;/<0<E<W<X<=ȺȵȯȨȖȵȵ{ȵtm h~Yh@ > h@ >6NH hVEmh@ >h {h@ >6mH sH h@ >mH sH h=/h@ >mH sH  h?sh@ > hh@ > h@ >NH h@ >6 hZh@ > h@ >6]h@ >hZ]hZ]hYh76NHhZ]hYh76hhYh7NHhhYh76] hhYh7hYh7)===)===8>:>;>L>N>>>> ?"?$?Q?R?k?m??.@@@A AAAAAAA3AUAAB$BtBBBBC~CCC(DDwhWDh@ >mHsH hh@ > h@ >6h@m,h@ >nH tH h@ >6nH tH h@ >nH tH h@ >nH tH h@m,h@ >6]nH tH h@m,h@ >nH tH  h@ >6] h=/h@ >h[h@ >6NH h@ >NHh@ >h[h@ >6H*h[h@ >6.X<O==>?@AAB2DD_EESFG HHIJK5MNN4OOPQQzR!!!!!!!!!!!!!!!!!!!!!!!!!!gdTggd@ >DEEEQE_EF;FKFLFFFFFF`GGG HHH HkHHHHHHLIMIgIIII*J+J@JTJVJ{J|JJJJ#K8K:KRKSKKKںڮڢڑڢڊҀҀh[h@ >6NH hbh@ > hh@ >hh@ >6H* h@ >6 heWLh@ >hh,h@ >6h@ >nH tH h[h@ >6H* h@ >NHh[h@ >6h@ > hR\h@ >hR\h@ >6NHhR\h@ >6hR\h@ >]2KKKKKK6LuLvLLMMMN NNNN7NFNHNKN\NmNnNNNN)O4OBOzO|OOOOPP麶顖rch@ >h@ >mHnH sHtH h@ >h@ >6mHnH sHtH hkEnH tH h,h@ >6nH tH h,h@ >nH tH h@ >nH tH  h+hTg hTgNHhTg hs1@hTg h@ >6]hM^h@ >6] hM^h@ >hM^h@ >\hM^h@ >6h@ >h[h@ >6NHh[h@ >6%PyPPQ&Q*Q,QPQQQQQQQQQRRRR_RzRRRRMSOSlSSSS T T TbTcTyT۾׾דׄyukg_gjhkijUhkijjhkij0JUhjh,h@ >nH tH h@ >nH tH  hma6] h@ >6] h]h@ > hsh@ >h~|h@ >6H* h@ >6NH h@ >6 hWDh@ >hU h@ >6NHhU h@ >6h@ > h7h@ >h@ >mHsHhrlh@ >mHsHh@ >h@ >mHsH#zRlS TTTVVWhLi]j^jrjsj|j}j~jjjjjjj!!v:!v:!v:!$h]hgds$h]hgd( $&`#$gd(&$a$gd(gdGgde%6gdmagd@ >yTzT{TTTTTTTTKULUUUVVwVxVVVVWWWWWWWXhhhhhLiMiiijjVjWj]j^jfjgjmjnjpjqjsjtjzj{j|jjjjۿۿ۽۵۱ۧۧ۱h/]0J%mHnHu hkij0J%jhkij0J%Uhq`hhkij6U h+4hkij hkij6 hkijNHjhkij0JUhkijh/]mHnHuh/]jhkijUjihkijU9 not. At the current stage, both approaches yield comparable results. For a more detailed discussion of the two different strategies, see Gamon et al. (2002b).  We extract a random sample of generated sentences. We take the first n sentences in the sample necessary to ensure 500 sentences that differ from the reference sentence.  These guidelines were originally intended for assessing the quality of machine translation, measuring fluency and transfer of semantic content from the source language. Evaluating the sentence realization component is conceptually a case of German-to-German translation. Amalgam PAGE 69 PAGE  jjjjjjjjjjjjjjjjjjjjjjjjgdmajjjhjhE[, 01h:p?s/ =!"#$%03001hP:p?s/ =!"#$%03 001hP:p?s/ =!"#$%0001hP:py/ =!"#$%0, 01h:p?s/ =!"#$%06 01h:p?s/ =!"#$%0x{DyK  _Toc11647534{DyK  _Toc11647534{DyK  _Toc11647535{DyK  _Toc11647535{DyK  _Toc11647536{DyK  _Toc11647536{DyK  _Toc11647537{DyK  _Toc11647537{DyK  _Toc11647538{DyK  _Toc11647538{DyK  _Toc11647539{DyK  _Toc11647539{DyK  _Toc11647540{DyK  _Toc11647540{DyK  _Toc11647541{DyK  _Toc11647541{DyK  _Toc11647542{DyK  _Toc11647542{DyK  _Toc11647543{DyK  _Toc11647543{DyK  _Toc11647544{DyK  _Toc11647544{DyK  _Toc11647545{DyK  _Toc11647545{DyK  _Toc11647546{DyK  _Toc11647546{DyK  _Toc11647547{DyK  _Toc11647547{DyK  _Toc11647548{DyK  _Toc11647548{DyK  _Toc11647549{DyK  _Toc11647549{DyK  _Toc11647550{DyK  _Toc11647550{DyK  _Toc11647551{DyK  _Toc11647551{DyK  _Toc11647552{DyK  _Toc11647552{DyK  _Toc11647553{DyK  _Toc11647553{DyK  _Toc11647554{DyK  _Toc11647554{DyK  _Toc11647555{DyK  _Toc11647555{DyK  _Toc11647556{DyK  _Toc11647556{DyK  _Toc11647557{DyK  _Toc11647557{DyK  _Toc11647558{DyK  _Toc11647558{DyK  _Toc11647559{DyK  _Toc11647559{DyK  _Toc11647560{DyK  _Toc11647560{DyK  _Toc11647561{DyK  _Toc11647561{DyK  _Toc11647562{DyK  _Toc11647562{DyK  _Toc11647563{DyK  _Toc11647563{DyK  _Toc11647564{DyK  _Toc11647564{DyK  _Toc11647565{DyK  _Toc11647565{DyK  _Toc11647566{DyK  _Toc11647566{DyK  _Toc11647567{DyK  _Toc11647567{DyK  _Toc11647568{DyK  _Toc11647568{DyK  _Toc11647569{DyK  _Toc11647569{DyK  _Toc11647570{DyK  _Toc11647570{DyK  _Toc11647571{DyK  _Toc11647571{DyK  _Toc11647572{DyK  _Toc11647572{DyK  _Toc11647573{DyK  _Toc11647573{DyK  _Toc11647574{DyK  _Toc11647574{DyK  _Toc11647575{DyK  _Toc11647575{DyK  _Toc11647576{DyK  _Toc11647576{DyK  _Toc11647577{DyK  _Toc11647577{DyK  _Toc11647578{DyK  _Toc11647578{DyK  _Toc11647579{DyK  _Toc11647579{DyK  _Toc11647580{DyK  _Toc11647580{DyK  _Toc11647581{DyK  _Toc11647581{DyK  _Toc11647582{DyK  _Toc11647582{DyK  _Toc11647583{DyK  _Toc11647583{DyK  _Toc11647584{DyK  _Toc11647584{DyK  _Toc11647585{DyK  _Toc11647585{DyK  _Toc11647586{DyK  _Toc11647586{DyK  _Toc11647587{DyK  _Toc11647587{DyK  _Toc11647588{DyK  _Toc11647588{DyK  _Toc11647589{DyK  _Toc11647589{DyK  _Toc11647590{DyK  _Toc11647590{DyK  _Toc11647591{DyK  _Toc11647591{DyK  _Toc11647592{DyK  _Toc11647592{DyK  _Toc11647593{DyK  _Toc11647593{DyK  _Toc11647594{DyK  _Toc11647594{DyK  _Toc11647595{DyK  _Toc11647595{DyK  _Toc11647596{DyK  _Toc11647596{DyK  _Toc11647597{DyK  _Toc11647597{DyK  _Toc11647598{DyK  _Toc11647598{DyK  _Toc11647599{DyK  _Toc11647599{DyK  _Toc11647600{DyK  _Toc11647600yDyK  _Ref8029913yDyK  _Ref8030558yDyK  _Ref8723579BDd D   !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~ Root Entry1 F0{G Data rjWordDocument00qObjectPool3ȌG0{G_1081603092FȌG͌GOle PRINTN CompObj/q  !"#$%&'()*+,-.045689:;<?CDEFGHIJMQRSTUVWX[_`abcdefimnopqrsuvwxyz{|}F' "   --%I---- $---- @ !F----$FLLFF-- .Times New RomanCwCw w0-"2 COther constituentsI3-"-33(3-3"Systemw@ CwCw w0--- @ !----$-- .-2 ?Prefield9"-"---- @ ! ----$     -- .-2 ?W Postfield93("---- @ !0----$-- .- 2 4Left>-"- .-2 IBracketdD"--3---- @ !0----$  -- .-2 49RightD33- .-2  BracketdD"--3---- @ !0K----$KGGKK-- .-2 4MiddleZ33- .-2 Field9---- @ !F----$FLLFF-- .-2 Verb positionsI-"333(33---%I ---- $    ----%II---- $*Ih*----%I---- $----%I---- $--' FMicrosoft Visio DrawingVISIO 6.0 ShapesVisio.Drawing.69qOh+'0@LXdpAc:\program files\visio\template\standard\new visio flowchart.vst՜.+,D՜.+,ObjInfo 1VisioDocument\VisioInformation" 2SummaryInformation( 3      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\_abcdefghijklmnopqrstuvwxyz{|}~Visio (TM) Drawing \^{ZVd !fffMMM3338 TZ Arial@F J"X F $Times@MNdZ New RomaUn@ NdWingdNPbdMonotype SortSymbol5T?? Y@-1U,J:DT1EW-h LT4* %U_b  0{Gz?R@&$6YEWb ooeoY ! k'(no RYY=,=,>p/&'$? AY&| ,,'#/5%&PF$&p o 3& y pnp)pL?^2 o o oooo*ooo廀?aBEEEEEEL@<`pBEUEEEEUEE@sO'FAkY;AVVA~AY??I\#25l =,=,=,#Ѓ}+cQb6W#Q Q Q iP| 'onW'07U> Y??4;jg2G#:}{ { J{ { { 0Rkjb#/L!3_۶۶D//b/@Q`_?*dۼ۷4 Ak@YUo%1߅-(Yp @ٻ !-0N`#ew)///ϡ/?2S9ϭ]op|7%nߤ$ -?Qcuw#U2qa?`r/ir b / "%)FX kU TY/}b /,>:/L/^/p//c'S@T#d:QGBT'a58R \RR_A_Ayb8Q8Qb\Q\QrQQْ+Q+Q ˑFqqFّّs AA@QBAFAARssFոF#Q#QFQQ@F=Q=QoIRJQFDWQWQFdQdQF ool}R~QFыQ"QFјQQFll1RQ벿QQ"@FQQ#21$T!RQ&#nA b ah"!)F'a"'a*Fhh+Fba[?m?????????O!O%BO$bAuoooooooo);"dU(`q} 1DC"-d!`̱͏ߏ@'9K"\k"`ß՟ /ADS"Šn'A˯ݯ@%7I[",|!Aӿ -?QLcϏ"\LAϷ#5GYkߏ"ߘ'Q+=Oas"',Q!3E!oC@.hH*ApC1);MWi{"fQ);M__q 1##Q //1/C/U/g/qH0' `Q// ??1?C?U?g?y???? ? mQ?OO'O9OKO]OoOOOOO 5O" O__,_>_P_b_t_____Ju )? _o"o4oFoXojo|oooooUo"*<N`r E" {#5GYk}ŏ׏O(:L^pʟܟMMM0BTfxயү䯤(%$&8J\nv///=O菴Nk#ȥ.@RdvψϚϬϾπֵ*.כ@Pbt߆ߘߪ߼ָQ8(S@r&8\*<$6HZGn+& C  .@º@ҿtlQ*27q ?/&\|?H  SA ki ĈOL$X(m@%>#'2%r"|n6޳?bj?-?X(X(SW7$?!3B0  FlB9kA!//*/Ybt_oϼρo& >I> 8%9UNlll bq)}D4\1{τ3 "<v>6B'#)1-BTfxt{7q/C6?g 3A6|<|<|7j(//VI!dMI/پ??/K]OO??b)2OO./@/ \/n///+2.d@/?)@,oS?hz*8OJO\OnOO"4OOj,uvVe?w3S4;::y __-_?_Q_c_+ is+(:[zwtzq@ybb m@|Q.//t 2aRk9G|/2xɔ`tx#\SOb(3isq,>os) G4֧{sqpy[uݒaCѿp_D_#*$ݒ*~ƅ CG!1ܶP WdX~LetterOasRIVA|'0  .} DisplayUFDfP 7h-RTUUUA@ ? I?d x ) bYbK))t!eqYkZUHn `f?p "6,Fx  { rޙ z|pj  "mAdjust the width ofbox to changeparagraph. Box's height as according text.Pmb?7i66?8j?5꿕HD '# =ih0>T@dE=7zA[U@?W?A?P} 4L >)A*A7PP4rUv݀@4u7` ?UB4*LVju 4b%J& !>74 U-"...$@4 @ rY?@8 %%\ M!u$`7-$%.=Uj#.A1M" 7w$',6$-%} "&bS26/+$ 4a BJE@{3I8A=D<+C`9Copyright 1995 Visio Corporation. All bBs reserved.`2`Shape.hlp!#18140l>]Udv6 KI2*(`3? //gR1 Q#R4R_/eTQ5___(gQ5kIVod_  0@ú=2ȿڸH%-!r  [Fu,#tB ԣNfa_to@+Lo6aoP+'-J1 UFDfP 7h-RTUUUA@ ? I?d x ) bYbK))t!eqYkZUH ??$6 3?9t zz?33npwqcUse to create either a straight or curved connector. Movetrol? handl4ge shape7 of.mb? ߿༚xV4ҿ8i6??5HDB &J# ?h4>TYY9 7AU@?@?P-DT!@A-4u `u bu ] 4-"44 Vg.AE D U4u7` ?UuUb2#4] tO A %*0"6,6'>?U"'///A,7?:%$#W<*G?4Qp d$b$ &a&$4 v 4 tNs2 6z2,3c&G@b6>@!m$2rq?@[Iz@?pDxEӿ??*brueLA@7ֻ0@Bu:@4B CC@:":"B0":"9(6Y`uA5@=@=@@u^`BQDB} 2:"W:!@@  V {3Q[Co`9Copyright 1995 Visio Corporation. All bs reserved.`Shape.hlp!#18150)l>$>Ud:!PE 30202a*c`a(:39`4i4KA[G  .T"yBKF%+%nvp4AhaOE\9 )G;z@'QA66y-Ue7Uc#7 Wq3i%$wW6Wd`CuP` Pwosi:aq(V![p7ti 3_'2rBq ?1+Hp-1r   $ضF|@#lKB tifao@+4o2a4:PUFDfP 7h-RTUUUA@ ? I?d x ) bYbK))t!eqYkZUH6 (W|~?ǟk | j f3??dj qlle_cUse to create either a straight or curved connector. Movetrol? handl4ge shape7 of.b?ԉ=8i6e?9EH D &# =ih<>TA$E=7UAU@?;fP!3|@A4u `u buU  -4 ">] @44u7` n?Zu[`b"4D4WV#4 / A h3AU&[ (".,.'>U0O?贁Nk?<[#4vZ$a',$\%x'@b !Ge#27q0??%6C!s#|v @m7\#?[m6Zy&Y6Z0.7M"2??4GO5'>OPJ+/"4B2"L2"{e#ADC`9Copyright 1995 Visio Corporation. All Rs reserved.!`Shape.hlp!#18150'-\#?2r2?@IP{UѦU??4rURS+*/LA.@bS7@jRu%0Tbv&kuvu4`u`@d# ݓ u`(>Ud[P5 e#{#a") " Y(j86@kqk)=x4tk\Y=4?!4Uq,#U=vZw&(nw92!6t<$txi4A^i\='3O`'%4w(,pfxkvS%?AA2}sVSTSHTvu"`d`Cu>6P Posi QH2"& 7 g`H4BU"q(\N9'TfMsN6![8aj'9Iv/;!3K0S IkGH-1r%  t+Fd#lswB \u fd\o@+ko1sD !P't4j@K@ rl&A-L r7"A-t4j@@ Lr&A-$371t4j@@ t'UA-37U()*-t4!%@& K'C-L\7"AU()*.t4!%@& 'WC-$3)7"AU()*/t4!%@& ̓'WC-,3)7"AU()*0t4!%@& \=(C-4N7"AU()*1t4!%@& (C-<7"AU()*2t4!%@& \(C-䓧7"AU()*3t4!%@& 3)C-)D7"AU()*4t4!%@& )C-|7"AU()*5t4!%@& )C-7"AU()*6t4!%@& <&*C-T77"AU()*7t4!%@& Tx*C-7"AU()*8t4!%@& d*WC-|3)7"AU()*9t4!%@& +C-+7"AU()*:t4!%@& dl+C-}7"AU()*;t4!%@& +C-|7"AU"()*t4!%@& ,C-S!7"AU()*<t4!%@& `,C-q7"A=>t4!@& , A- 37;?t4!%@& -A-'7"A@t4!%@& C-A-'I7"AA%t4K!@& -A-7U()*Bt4!%@& \-C-7"AU()*Ct4!%@& ̮.C-/7"AD%t4K!@& p.A-3v7U()*Et4!%@& \.C-$7"AFt4!%@& /A-,' 7"AGt4!%@& J/A-4'P7"AHt4!%@& /A-'7"AIt4!%@& /A-,'7"AJt4!%@& Թ0A-'7"AKt4!%@& V0A-'\7"ALt4!%@& 0A-7"AMt4!%@& 0A-7"ANt4!%@& 1A-#7"AOt4!%@& D_1A- 'e7"APt4!%@& 1A-t'7"AQt4!%@& ,1A-'7"A)@t&BR@D&?R@ '>R@D^'?V@'=V@(=V@P(?V@(?V@(?V@F)<V@)?V@t)?V@9*?V@*?V@ *=V@-+?V@+?V@$+?V@̦#,=V@|s,?V@$,>V@-;V@4K-;V@-:V@-?V@<1.?V@|.9V@.?V@4/;V@R/;V@$/;V@/;V@t0;V@^0;V@0:V@d0:V@%1:V@|g1;V@1;V@1;VH<( H<( H<( XE&5 REdv35 REv@5 R_(+<W'B!<"/"č%Pi{ /ASew//+/=/O/a/s//@//////?D(2TD$M5dP0Flowchart Template@Copyright 1994 Shapeware Corporation. All s reserved.QA VISIO template used to creflowchartsing the shapes fromrt?.vss.v3.0h746C96FC;6SC:TG7AVJV ?U ?g "*^p? N(~*~<7CUP (?U+?FDTey ahm[ qT^UIj{@I@2[?I?*CUgCq! qb*2@?,Wu`J5";"SuaE)eH0 |$) J;!!#!(!'!(!'!(!'!(!'('!(;!'!(%#!('!(*(*;!(J&!4JDb"5ZA!2gA!3tA!16A!7A8A;!9A!:Ac;A1< AA=A,A5@O__!$"ik_ ?Q?Q!?QUaNQdQE?A?AO ' d 'P eC,h#0O贁Nk2 &gb^bzZe?@e?)qp!ah zt@@"boop!Havo~UA.  < tNC !zsazr r!sKT2v.HR"NQrv vTXogA~?@` %k?#?.gla$q2% <UjU!Ao#iF$`Ao5e#2n_p^u  /2!qvn!5jxiTAtE'2>b !!`B]`8`_r_]Vew_g@?@Hz_G?Ф1䭵= qݑC x! jBsB1>BqU@ьPr_?@OʀމźȺKW2__]Uh!զ8հH߰:Өu!߁uߤTO2-?!dej|օy @F% !u,AOther constitueGntsܒ ,5G&x K oBT@m<B^2$>*w9KCuI?Q/Aԧĭ!!5+=O ase$e{Կѿ]BXx \X(@?8܉#̛ϭϿ+=?-_ߨ?~??POoI|']o暑⡹PrefieldG?0TC. + xRp/_1oU)oao @,OoO+=Oas}q //w@VU3 @6/@H/Z/l/~/)p//z/ ??1?\_U?g?O?????? O-O?GuTcOsqOOOOOi|Yost%_7_I_Fm__oo___c*oʿrooo|oo&8J\nёG,/!j5%BTfx!Qc@bʏe#n l *XOjO|OO&OOJ M __._@_pd_v______Ŀ_o BHOZOJ OOw OOb O/;M_qvQ}V"vQ1}Z2i{<U@VU`@AQ?AOǽdP݀,u[տ󅍨# 30O贁NkDSRrrAVd?d eL )Qpew@z300"_rewRpΡHqv~(| 0ؤ3R- tNM`Ρzsq- 3q@vJO\OnOOOOOOOO_{a-_?_ae_jo|o__#ook=oOooooooU6234U5789U:;<= t4 j@^@ wG'C- HA@\HKRH<( EH R\|\8tG@?fH.TD,H.PPage-1Black fillWhite fillBlue fillRed fillGreen fillCyan fillYellow fillMagenta fillGray fill10% Gray fillWhite line30% Gray fillLong dashed lineBlue dark fillCyan dark fillGreen dark fillMagenta dark? fillRed dark fillYellow dark fill50% Gray fill70% Gray fill90% Gray fill1pxl line3pxl line9pxl lineHairlineShort dashed? lineArial centeredArial top leftArial topTimes centeredTimes topTimes top leftConnectorFlow NormalFlow 8pt centered Conn. arrow endFlow connector textAuto-height boxGuideGesture FormatLine connectorLine-curve connectorAuto-height box.18Auto-height box.19Line connect?or.16Auto-height box.17Auto-height box.20Auto-height box.21Auto-height box.22Auto-height box.23Line connect?or.24Line connect?or.25Line connect?or.26Line connect?or.27T)RR3lw7I E3|wBIE3wQIE3w`IE3wnI E3w{IE3wIE3 xIE3$xIE3?@ABCDEFGHIJKLMNOPQUUU U UUUUU !"#U$%&( t4'j@.@ dUC-̂ VA,%t4 5_bVA-,1h7"A@lvVJR@TNjV3RU1( UO"D&aU=%Qb )h"Ty+U-xH="(':*EQ*&Q),$H<( H<( E$wX RErX R{' +* g"4FX (h([@(0?@(ϊgE#Ȏyz T}[:TB|V  W&L\x!5>Xc&3YmF}`wK3o0F߱5V$ܮy6wT$N7xNRGƷ)G48 D]?Hh2|\N,?VVT{=!XR=/1 !=5 `=D5 &!PDocumentSummaryInformation87h_1070196832'F͌G GOle =EPRINT`Po (0 8 p  PagesMastersPage-1Auto-height boxLine connectorLine-curve connector\(DLT_VPID_PREVIEWS   FMicrosoft Visio DrawingVISIO 6.0 ShapesVisio.Drawing.69q  s *A? ?s"*`?2n.Tvv*AOiJB`!B.Tvv*AOiF p:7p6xڭVMLA~3]* FTј`ҋ!4xCF]x41 d\ oxFO\xABz 1QL7](hv_ggof;C ^s?#:&h؂QQ'LDDO GM7()k%IF!NdzuGE 22`+ S`L<"˺ɰSƭzy;CaTezci9A:Y\%e\jR3F!Β#j ϊ|O!{ѫV.ڗ]sO_r2\.Mi5gDңYvXN6!2Ġ];*1=-ZL5&44K>JҶ5N{&dU֠Vqm?!S n{]dR5h[ép 2{W[_F#ό0Yqmt֑Bܝk X* Z*-e,B55g'i :mTOy L}Iu˵.ߒZ]Ղ4YY}oȡv*SUاVxRG b)z[9eeFY۱Ş_L4֕珟Im''uccbɢ ծJwtF}N<wC_,uQVXT^i=H_b~*>[ls߷0V5քdX.Ӈ=\} srӌ;=|Ѧ~.]{g㿟?WEtk7 cvpyw{]y<{* 7|aH9bǬ]C?zܶq'#"Bw~BzLv&b7/xrz[}^$cLOx}<M?3A7,3w6ПO:sͳ=q̩=w}Aܟ|?(O} v\SXܓd9Yר + srૌӲf_E9З +3pLfR˸+Ky?!\П㥤?3]M:ϟp7JRuu \:¼ ڜ$&\3?eF:/v7):֤d&Oѱa}2S.ypݾ f~ʬ&tUWZ\7z=|)s:VkᬸX?eNx:?+Sj4Kx­|-7(";p͙> zwᗋUr~-yuJw5317p?D]\Ϻ\=nUԇkSbm9b}lIw.g8~"Zܯ3U㪹z^.9w'f7>wC\pfb9g]yp.Z7{53j}FNj;T{\n<ڭj0wkUz31SdeLqawp)5<n^}K"X8m\9qwFS#gTWw "9So epiա[wڏ9} lS_M3AB6ғ,N1%ջ)?F¥x5w?8 4e]OW]?z0s|rf"wh:uӺE.}(\d^w7Ի4"8gʬ<Hq`bB߁"j, '0oi܃.c(1'o"̎} 's֜3o+33 }lFvˌ"`MIj],6Kvn3Y.`qkeƳzGS%qWr(g"Ϝ=8>d] 7*qEM 9)UogqMmV?A9k 7yZ^muVŲ*IU_[N+ipˤLL QDX[Dȅ8|m.{ڜ'\dg;y}/U$[Vn㮌keuկNғbsqs[Xk?ec~h= pqu?exָqކRwq6|pܕqp׊I:ӯg7X7{9Q8a~8iͻk88na]_EU( y\nn1ޙ.FvQ[7+c.FE:}`;nQ .dO|@_uil`rX?LJk|?08k+3m!K{VF_e[ W#݆ƨ1&=nZڨp/ =>n ] 3W 3GC,q+Y|Q{¸^5nGu̞y| pϟ.ܪq1i(gCb (r3+u9{򪑰Ib BF YT]\LU.)l4A?1]IV0Q9eFM9 Klq\fxrLs"x`'xQ1^e:N$c`4"OF5G.xh } .r9|E<,/ ;^~^3lF0`F`F0~3FUStDk/K*LwH; y& 8/wqyL7Eo*<噱 g !7Foq4ClYઁyو'T}l𞊡S[ I]p7.Z?ei}N\j8Ecx:m111Њauяh VϋΫ`F`F`b|ȇ]<_c/|lv|=J_̯h1NJ;\8wSW[).FO0`F`#0#0.`|8H{_'h)-̫;8˹x`Qϫz7l+z< eQq7:oa$JiMG;gF0#0€a0qg2 nfQGVM=lHש|C16Ou=lVcJuֳ4oӡawrol`41 )0 5o4hgFqިw3(]u_t1Qˁ1:~:w Yu\ԝlY8aWa JN i[e>)oݜowh:MQVYwvmغRg;lbo\N3h[LgV^7h7NO H~FReyzL֢t #o{T!}f/Z mG΍p^)h"o#0#0€ؾR/(cd8c)>Z,58b 7̫$q7zqD3sʲl.<yFoQ=-=v-{f|?aţ?ںr+>[Vf,܁?S30#0# ƆN8`K14ijq Z Pq Y#,AYoro\lA#g}IOuL cm: F 񉉲Ȱ#>\|qZXG9A8uMO:0€6ToWW|FmƱuWaF \W|*F+Ɛop]ިSod_qˬQ+>#C|b_YW|7€8oF`F˔Y/q8^cˇ+7kcsy(71?;~uD7'FesS/=Atv[t~nj3Mms%ƻowkBg8*x#?7f, :c?4ycD {Ol.zԶYZ3 ;rm7m0`F`Fj@Z__voWpoVa%.v6,#鍴t/q`nQ505vM0f5t3onf7Am8- euҨKF5:bn{#?GhZd~u 14xsM;h o]`n YUu#g5d_ 6F{cco_>s90ʎ˝mr/_0`F`#0#0.`w0ZfsDxAkSWFߨ@sE$?y#0>yVh}\N8760`F`F|()Y"iZ|: {p)5$1V9iMix ,Mk7Ŀ #=S<)o,0Rt㩷{0y|d6Ne7VEi}VxKxA-g0o#Vn{cz;;cZ2]`-=l}*(.t1i4-k~];1Sw´ { #ϖui|huQo/퍘/YIkd40b7.1 Tm/YG< Y0`F`F0 | ` o\<` F7F-bcwjXQ7ڸƅkJ cdiݮc%F x1Q_h.xh<5[vhԗQk4a]s0`Ʒ(Cmj>$HT~|Q 9I1ڤ[ZWWxЌc%)ѥ[}b4&;:jBשҿ /`ԞF?x* fWQx=,tLl[NdN3u.fGdWaOF`F`#0NQ*su=~t ǻ-suFǏF8,vr0."q懼ЬUMzx 2Z KZ2ƣ(6w︶4cF- 5s#acrh||0soQ6F!1 ubcZyjwcNrJ .ϱݾԵǚbh/V17€[#\CƗxU qs I1d I|/ޘiNjKV%Io5cOCz@e Qʢ0o0y~{-BkvZKqNF`U q97ʤ {:b[o4u;1:b[ot1ZOGl)yn}[S70cOG68#0z#0aذGjo 鍏^7]Ao,nS7&t5 qq69!LGܝ(CdWk?z#Ҹ^]MN٨uP"nS7&Ø鈒R4nWk3zc7\W̍潅 h-b<5 F`F`#0# %1?S1b~g`_(4~7b~W`ԟS?# 8:I@X_Zo02S_ES9ƗO]pwO]p `πqp7#0#Q-Bg)4*QBQNpnXB\.Q-_h`lb\xo\(1o,Ko"%s\a9F(ҘnXD\bDShDtF]5w1F`F`#0#0€blWnIENDB`}DyK _Ref531749592Dd<| 11r  C .Alf_exampleC"bTS/v&12یϕP{nTS/v&12یϕPPNG  IHDR%?sRGBPLTE?$7IDATxiFZ+#n1^lhB:#:uI@elHr2]0\.L@Pw\>G|#f%kL%V]`q#\Ri7+X!9`8U0o" +$l V$!a7T\Jn1^?zRP.Z2t5ݢu]I[o]HUAN:GQ70xfLV̉Rh_?¸F A($r7 # XCƗ=@@l]&@9EA`J9qZORSwA7 ]{{6&u ҤBF@NǨI?X`[FBB@Mt&g^C@> "6OLɝ*dV? @DZ$>%@hp?~귚4ݼ*Œ)BvdVΦx]M 剫 Jo;) x2}av-\\;_\pXy%7yo/+^*]',W:؇`?ōW2 `4єs@^o!mTѰG/gfkzZ.]-NeԘ!ץ=՚Jߘ u$} tV.MGI<4#ߠ 8 %ԇG@N"8#\0҅q } bY$De(ɇ%#;FU(*DbCE`af/%%>+#`"OI+3o>wM >Ndj0 8odrSK&`7W %d5JJ`;w>8ۿiŝMψ nc=wX  r4<~*ňJqfUU O%Mͨm~t)#)i6')4}#ىhNP nV`pЋ/'':G_ J ֘l%D"(B =`Nw'&@K!`kSlD ;LzR2 b5k,e>@6iE % 2kC %LֆԏZ C4kӏn V|m8p'1 aZg)²TJ@Cr7ITW9ţu!&y}*&m\;qM@E):Zy22t^Ac j,hNߪ4Op>XHd\ TSt`UN l0gFӃS}>GB# ENٕD`+aǕgYa\!ujL@E *CA#*t[ rp,,<;Ȏ';Pg'kB19y*3IĆ[W {Cd,hBNDa:m ֖Dd{!ȥGT.B&HX- 7%ÖB-̮휅DXJRtr;\Ǥ`sKjLUJ>IL@(2elh0N'@HS}E~ ܛL#1&ZM0Wrvː,OO.@j)|r"6x`B) ܍Jp튀+!M6Ұ%,"vQ  &[IO@F@ش? w(*냴~>y }* li U3N0L@C>@m䯗><΃DMP "%H[΍SQ:6O=;zuJчuP}A_ZȣL u:nH$zA BSPH)C s)E'}AD$WX~+2,.Ӭ鈀KCF/%H&YwgCo K@Lؕ, LsA˔ @@Y)"mN`B "fSٯy4X2`Hn|n@A`L@F^@(^c@X:Ș LS_au<`75<eN(00k7uG3RGR`f:%uoHJdn9bҲ4Ӎ, Ll L]V{엀'JE {Lx鵐_9z!@؈ 8"ғ]?O\ֳ L_Dk ) q "V,Pk<>蘀굞o41b~^bq<)Tek#@*_DX&!, %4 LV2HuyzF~-GdֹDUp w*Hԭ\Vޛ, BY9a-IO 2$ joQ8 ;\A7z,isY$@ - ZXHĕ"r%(M@>U6_ &`Á\s>@S$ 4ҫë ѤBQ <b,c>@{>`*:Os[i50hbbZ٩+~ 퓣V4џIvo(h|&4>T hXGO 2ʖcoaմ/ s[ʸL|iG >ImM7◕.$p 2~UY[FEr.ZyMe H<M慁 v}`$ݙAa@dyYU/5)n<\>@j{MrYKmʷ M8el` G h€، 9Ka@EC/0$&0AaueGf)\b}r 6z8v0-`50+1ȸCHgezPvSL 5#`>6Qi>p|Nd>GCL!Di#=@jt`>FV ">=x|`,0'G?` PԚ"P<fN\}4PԶPYEE3Àһ4JOL2!`>՟`gz!3I@vTpE"b-\$!\~].P65pY25U슀0g_^ V.cc~Xv]Œ&T+E/S-'ЩX[89I8B > 89<|r8;WtL Б\ttL^]*HD+C&2.4 >(W<pЇJ4op@"  rJc,^[(7|.LX#r 0\.L&  !{GheIENDB`}DyK _Ref531749592{DyK  _Ref11047897K$$If!vh5"#v":Vl t65T}DyK _Ref530829785}DyK _Ref530829811}DyK _Ref530882302}DyK _Ref530882316}DyK _Ref531749173 Ddq%t  S *A ?inputS"`b MRӿ&+S^u Wn MRӿ&+S^uPNG  IHDR sRGBPLTE?;~I IDATxKi>_p8Qc??U$ 7aKv!X%+U' , fxK J?GCɯiA5+/ru[(0.zXEP?xeYfJ%$ߪWU[Wfz3,Q=N#R#pVg|20Nfua]S֣֡e#>P _/3@uzD 󤧱6XLHW<0ͅuMYka-@6 [E<5I!+ )DvQAi:"T˭wc N&@ȏ'`S9 n4.p[ PfZy#,@}g r_8*3'O"b樎ID6t9'6Qֽa^6mׅ ]&I*u2w _wl~`I2g_A@h۪rܹ/$hb(/L |^thQn)ҍJc %]Uujm]`I', V4Yu0!*C _75M#&4C\SrXmr=Q_9E#wZ EfhoۆtJsj jUzt)nbX#Z %; >41 mMp,Ј y!+u<}!TՐ=. Ъ1FjUbswubVc ky @<{`΅_J1a0vk@ J&uYJy=P9Mp`FH cIwaIO6h)@PxO"=I#V ?\ i>H3C\%NJ_K86-.mUpc-+?J?#ҏC|-ո괎|< 6[Ƕ-w|^~xl[Z_q뿠? o.dlc -#OX村fx5ͬqLL,px 8/ssmP&Xe2 T@;L-px+qǃŽ?aUUxוt#mЯox,PTnǎCi4@ P~pG ܏n ~⏗7nu}6$x0ݢϑvH1QfN ǵZ35 9冘:bz~"'T{ ~w-CWpc#4Tm.:*q o:*E@Ul֛ UV;| nHn,l)`bOauc .@@<?FP0MJ]onBiطx:&) Y G߆dCQTl}r(4_H܋Un^N| p%CF}=w}{3C%{߈;?u!]:[xe{OGwpqA\?gP-<!Z>[84b\rB!"6g!<|f-:-0~U'}8.&?;Au?`qh3% ҇w9w⃲cExj6 @#bAghCN8Zk|2n?P5O7-pi~۶@>`ּxs:3i9{P+9C8} xC}PsOz-i_z~D᝭.xg'W,c}HCV8OhD}lN \> S<@SO/ӅE0Zy*!:[`m[gX, ,\_e.<*;(t1\WhUD̷XO[`q)>XJ#n~'$0 T5 އXRX`I`qۢ HOIENDB`}DyK _Ref530883055}DyK _Ref530883055}DyK _Ref530883055XDd5v  S DA&?afterpreprocessingS"`bv.7?;iLzk<RPnJ.7?;iLzkP :J°h2sAK{98ܕ;A)V'ʇˠOUP'"(ayr4MRBV.` \ǴL7&nm,,jJYY ;Be:ȍ`v\K($^0kl~RNg YPȫ)[ʨ|boCOI gcm"ʠa¤lnp5 e# ;dI*wRxxS2N`:h.I:֥B]5 hW׀c@Nু?raiJk:m3v7i<6ݒvQ2] 7(.JRXr rNRMւV0ӷgUvu:&$(s]Ә8=-IwSma3KTM4KeK!zIU1XGC^/b*PBj!eVv,90-X2*'JZ:I޺P{mQvK5QPك#*Æ=T(η2vb it=Թff mAuLjw3=0湁IP2)F3<[A,+  ̯ Vڐ_cc*H'zlż4FdNA=Pa+ܱC)V9X}(bաB(^Z|gWOf˄Lds48H\Z4g%ɱP(W Y^z2ǹ{HZ0,̎iJ~2(2vCQ^c+x9lrU Z%03(n1qM (SS=z}z]}80Sz%RW,|G S4M:?]4ɔ!h&/2*#(ps^k ȴt!jkf4wFD8b'2k*![4L;ss@i ۅV=(;بW&|سaQfBfkQpg@rG+a3+0╆VIuBƴ5M3|fKE!i(?Xagl~U)%(؟@)3iy׏rk)~A}(cl,omr92M%#s,xe 5*cB4>;ŕZx?C9].6J׳ EMW\YU;Zڠ}Sw]RO}ϸTě eSDIZe({_\YU~r`?7koagh{|L5,*i 7IcӧkW7#Om'͞RLP2J5RtˠWAi]2(S|*X7~ef=X/r_9LDi;C Pvh'E8Gu|n'>y@P:Mx˺ fWN:g epʧkzY6`N͙&eڠkڢ(NfdkzM)Gr-512kj]\/3K2{7Qeµr(MÞ8dS6 q(r(QV(C0W6nȣm!Kf[nBD"ʍ&(OC)7P%(~r!%JPJ{!Ìi72u/}tZcP.^9?:N1Ӊq^7QDr=LO[}/ǭ{iX:GeBxf'K(u/e,$(_ə]2sc-RP(ÄVi9J*w^%ALeG([η[ \79xK }vG+(ʷMʃlt?D)V!JM.t*2( ('~A0[]2dX[dyIPc![Bz}zvn[Ol(j٣Ϛ a'ɷS4OZZ sk T 5i[tHS&T9Ay=5Oc`:ɂ-[Q'uapŪf|Ik 1eu2<}~&3.(SLZĥT?e)AB|L;u3u*3:[S!cDԗ:E$oFD K 6iZ(@}Kk@91lr(v}.;0K:i%1je2_K3@ke4! ive-u2<ʼ-LRwXB(}ZE(ap8{fĭ ?S֥ Tc|NΚJeaa-M_AwEI=0t oi(?7N9Ͼa(c\s wꜯrb"ʓfدKP JAyJ_ <2I+m2+6ִJYIJ,s(ہLq/X&5~MB$miM4H eeQ>'W$l0%M4$ i(Z1 qM (S S=zz䘴loeHbZm$ i VI Gۤp0}Iѓ ՎYCQBt M9[I@Fڃ.iTC\*PڊP0(lUȼUbbIdLo&秄e:(;بC&y9GY&aϺ6Fk -8a^%wږs]?aJ$^&U ҘUϬ8 zTVEQ2kX^MgU܌i~ZÉbRP JA)V'Jy:(]$Lq16? P鵕5'qc%( hhR_9V҅lif~nBm~磴 (YssTJ(YK7u_X-!C)~+@#UjU)\+ϱ ,J9VgQq*s 3A)(RLP}KچLIENDB`}DyK _Ref530884156}DyK _Ref530884156Dd5n rx  C 4AafterfleshoutC"bաs DTnաs DTPNG  IHDR9}CsRGBPLTE? "IDATxQ, .xE_HvM:3uT^*vfە3|a͙v9\EP.]Q6R7`f* s0KXu .2 i,Gh,:~fڡ_2ee糽]6ixt]aFYR:%`6aHy`ckek7i/7:yl|6HxҰuFY:B qL6vofje0`#uޟfB ?.U-"V L*Ηlx$鬳2fj6”-3fy`Zfer˴Ga#!dlطFy3t`nV4nOuҡ.].Њ 9^7(g{.܍ qq6а ƚT6J*]hHebݟڌsw[۷0?cs s3j S0Oz<̫<8Lw쫪əa@uOݖu4gɍftάۗ LBAg Cr_JuNII~O|Sx)U(&V˪2lPy {6]aZwڬw* lmtOʆ *"af0A|p%Q0KDe˸.ݿ&4s.COgCni[bIJiܷب"ļ)Ws0{g|onn4NWU.^0@k[u.ie|P9ilKp!I:ۍp4nm+3}ezj.Љɻ!p%Hsgj0}LZTp62.o30c^c=,Loc@OLyO }5 -ZfG-ϝ[f f;3?gf0qn/wEwa409z@ai9L_0 %TR4u/zzuU.ItOjD]~vA 4@Cc4ch|+c/c&'3=fT_ wU34ݹYtK%Olڅ Vo>/ D/?q>Βej8^ 5ΩgN͓ Ƨ^5Jbʿ'A5A=ӟ( Ƀ fÒ5ے~hq !_2-tIL}A0 `2x"Z&w4(K\$y櫺f;0'@ gT* ]oIߜk 0_5E^؃3/I>eNa:5]oq ;`roFߔZ̾7T+Lo49kߧk"̕2?̫<L=Ӫ`ftuC _:E]()7`^Y`Y::~(q^8_O$?ξ:jk9N:]̞/8'G3qꨰT6,Yw7aH)+w4BD~kfBsF}LPy0-=4Tǖi_gGgʁ-[9Ԗ70{ FlI`>4pkWUrб`=f'.{4,[OZD|ia?ӄ2ܷsCôO-nN>-o:'Ü=o´fM09$N<q~r~N#z|膣0Y˴ 3of-^nvλ0??iS2} i9ǹ<fsd{_2_9g dT '=ǁq ǁyU?#|ȏw}A $<Saan{:|ȏ4AS>tzďZSa~ }ď3L0?3ܛԏqڧ<~vxTT'>}*̏<6oAݸ.4?m02T%Lg07M(Q`|U1imL(m=|)`^|zTN&vm:b'2?.<.L?' }_oMqh_LLډ:ٚG fI\w4|>Ψ_VE?N[3CL Ijnw_2<es1i/+*?0j3Ĥu 0pi{eugh_\L&HH}g)0`mW@sSSM0iWG퇁T)L=i &fAjJKA,L|YnX@NL\6s|1#cj0(~A&VQ/WI׾ln ՗L/㰗sǔ0I&bǂCiqx/ŨM\Ů u|p`SD(1X/ 098J `c08i 5ФMV,=-oI-3zM̹49B0/2Ȓ" 1 ]jIц!aFc]ʟZfI[Ob%L̏ g*s`1L0C7uO430 BGz430c70c<ھj퉻:[b5ƒǦb0-괆 e'†G_feY0վ ?̫<Lɾ4.`a'iCܸd2OVX̊,3۟I/;m9ѳKa9nMe($,>O-ߘºAW mw0QL/Kn : WӺy s3tƄBf؆%}Sx JE FKC\ 8P'P999L}H94g6 e<4 L˹ц旞Id`fgbi1l AlBoC^r MMbicsq9q6̧@R,3e0='ۏ0sf3iܝq0ɯnf93_n307?̬ .Lw1`t'ţB vVN fh\,rOasqzm,4]Z7;]:_v =̧-?aPXrXA|>߃&}0u]a)L50 zacрk%]үe07oIQE954`Q k A;XӜ%uv^m!삿&LZc,DۛaI07&ƘmgM/畺$0=X7 ^`=M8)}ö#)Ӯ%4s0-sy`qɼ`Y0_;CdiCF>99Űuͨ %i(pA2])`d      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~]li4p<pc EMFPo@F, EMF+@``FxEMF+0@?@ @ @@< @$V@9@13KC"A!b $$=='% % LdKP !??% % $$AAFEMF+@<0NnC@@H<V@&BOC&BOC9@V@9@V@&B@$$==_8881% % V0#K  PKPK% % $$AA( FD8EMF+*@$33B%BV@9@@0$9=ARIAL6@0. Pre-Processing?""">?""">Ǒ?""">x'?""">q:?""">E?""">UUU?""">9^?""">q?""">|?""">r?""">U?""">9?""">מ?""">?""">?""">c?""">??   RpArialsMonotype:Arial Bold:Versik%`h6`h`C`h4`N`hB`X'<`4w=H]3 `$w Uwww w<(`^3 `p`3`0dv% T; AA;Lp0. Pre-Processing% F@4EMF++@ @$I% % $$AA( Fl`EMF+*@$33B%BI""">>""">>""">*>""">r?""">r\?""">9.#?""">xJL?""">]?""">rh?""">9x?""">?""">?""">0?""">?""">?""">?""">Ӳ?""">r?""">?""">??   % T%3AA1Lxamalgam_preprocess_lf  % F@4EMF++@ @$XC)'C4C(A$$==% % Ldey O !??% % $$AAFEMF+@<0Ne>@H<XCBCDBCD)'CXC)'CXCBC@$$==_888% % V0h* d&* d&y y * % % $$AA( FL@EMF+*@$33B%BXC)'C@0$9=ARIAL6@simplify_compounds9N>""">>""">c>""">c?""">U5?""">%?""">+?""">3?""">A?""">Q?""">`?""">Uo?""">Ã?""">r?""">U?""">9~?""">g?""">P?""">??   ( RpArialMonotype:Arial Regular:Vek%`h6`h`C`h4 hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% T>AALpsimplify_compounds  % F@4EMF++@ @$XCB4C$A$$==% % LdqeO !??% % $$AAFEMF+@<0NnC@@H<XCDi CDDi CDBXCBXCDi C@$$==_8881% % V0nid&d&% % $$AA( F0$EMF+*@$33B%BXCB@0$9=ARIAL6@create_lexnodes9>""">X?""">j?""">r<?""">9/?""">8?""">XH?""">*X?""">`?""">o?""">U?""">?""">:?""">?""">ӡ?""">??   ( RpArialsMonotype:Arial Bold:Versik%`h6`h`C`h4 hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% Tv5AALlcreate_lexnodes% F@4EMF++@ @$XCYlB4C&A$$==% % Ld;eUO !??% % $$AAFEMF+@<0NnC@@H<XCrBDrBDYlBXCYlBXCrB@$$==_8881% % V08iYdd&dd&d% % $$AA( FEMF+*@$33B%BXCYlB@0$9=ARIAL6@ DeepDegraphLFQ>"""> ?""">U?""">W*?""">;?""">AP?""">`?""">sq?""">U|?""">+?""">ێ?""">?""">;?""">??   % T@3NAAL LhDeepDegraphLF% FL@EMF++@ @ @$A @$@>!b !% '% % Ld!??% % ( % " FEMF++@ @ 4@ @<0NnC@@H<V@xCOCxCOC]CV@]CV@xC@$$==_8881% % W0K  K K% % $$AA( FEMF+*@$33B%BV@]C@0$9=ARIAL6@ I. Flesh-Out1?""">99?""">UA?""">r|I?""">rZ?""">b?""">Ur?""">3?""">?""">?""">?""">`?""">??   % TKAAK LdI. Flesh-Out % F@4EMF++@ @$I""">j?""">r<?""">r0?""">U:?""">I?""">Y?""">1c?""">k?""">z?""">9^?""">Q?""">:?""">?""">ӡ?""">??   % Tj AA LlSyntactic_label% F@4EMF++@ @$I""">DZ>""">q>"""> ?""">r?""">9'?""">g1?""">8A?""">R?""">jb?""">k?""">U{?""">c?""">?""">r?""">r?""">U?""">9?""">??   % T4sBAA@Lpinsert_determiners % F@4EMF++@ @$I""">`>"""> ?""">?""">(?""">U3?""">9>=?""">M?""">\?""">An?""">~?""">U?""">?""">r?""">U͒?""">9V?""">I?""">2?""">??   % TjnxAAvLpInsert_auxiliaries% F@4EMF++@ @$I""">r{>""">9N>""">>""">U>""">>""">UU?""">'?""">#?""">3?""">*E?""">M?""">\?""">r\f?""">Cn?""">U~?""">?""">r܎?""">UŖ?""">Uu?""">U%?""">?""">?""">?""">?""">??   % TyAALinsert_expletive_subjects% F@4EMF++@ @$I,6>,X4 >,I,V,d, o,o,6o,X4o,d,V,I,>,>,=]>,6>,X4>,I,V,d,o,o,6]o,X4Oo,Dd,DV,DI,O>,]>,=>,6 >,X4->,8I,8V,8d,-o, o,6o,X4o,d,V,I,>,>,=>,6>,X4>,I,V,d,o,o,6o,X4to,id,iV,iI,t>,>,=>,6D>,X4R>,]I,]V,]d,Ro,Do,6o,X4o,d,V,I,>,>,=>,6>,X4>,I,V,d,o,o,6o,X4o,d,V,I,>,>,=8>,6i>,X4v>,I,V,d,vo,io,68o,X4+o, d, V, I,+>,8>,=>,6>,X4 >,I,V,d, o,o,6o,X4o,d,V,I,>,>,=]>,6>,X4>,I,V,d,o,o,6]o,X4Oo,Dd,DV,DI,O>,]>,=>,6 >,X4->,8I,8V,8d,-o, o,6o,X4o,d,V,I,>,>,=>,6>,X4>,I,V,d,o,o,6o,X4to,id,iV,iI,t>,>,=>,6E>,X4R>,]I,]V,]d,Ro,Eo,6o,X4o,d,V,I,>,>,=>,6>,X4>,I,V,d,o,o,6o,X4o,d,V,I,>,>,=8>,6i>,X4w>,I,V,d,wo,io,68o,X4+o, d, V, I,+>,8>,=>,6>,X4 >,I,V,d, o,o,6o,X4o,d,V,I,>,>,=]>,6>,X4>,I,V,d,o,o,6]o,X4Po,Ed,EV,EI,P>,]>,=>,6 >,X4.>,8I,8V,8d,.o, o,6o,X4o,d,V,I,>,>,=>,6>,X4>,I,V,d,o,o,6o,X4to,id,iV,iI,t>,>,=>,6E>,X4R>,]I,]V,]d,Ro,Eo,6o,X4o,d,V,I,>,>,=>,6>,X4>,I,V,d,o,o,6o,X4o,d,V,I,>,>,=,6+X4++++++6,X4,,,,,,=o+6?+X41+&+&+&+1+?+6o+X4}++++}+o+=*6*X4******6*X4******=*6Y*X4L*A*A*A*L*Y*6*X4******=*6*X4******6*X4******=e*64*X4'****'*4*6e*X4s*~*~*~*s*e*=*6*X4******6*X4******=A*6*X4******6A*X4N*Y*Y*Y*N*A*=*6}*X4p*e*e*e*p*}*6*X4******=*6*X4******6*X4)*4*4*4*)**=*6Y*X4K*@*@*@*K*Y*6*X4******=*6*X4******6*X4******=e*64*X4'****'*4*6e*X4r*}*}*}*r*e*=*6*X4******6*X4******=@*6*X4******6@*X4N*Y*Y*Y*N*@*=*6}*X4p*e*e*e*p*}*6*X4******=*6*X4******6*X4)*4*4*4*)**=*6Y*X4K*@*@*@*K*Y*6*X4******=*6*X4******6*X4******=e*64*X4'****'*4*6e*X4r*}*}*}*r*e*=*6*X4******6*X4******=@*6*X4******6@*X4N*Y*Y*Y*N*@*=*6*X4*|*|*|***6*X4******=<? % % $$AA( Fl`EMF+*@$33B%BI""">9>""">Ǒ>""">U5>""">>"""> ?""">A?""">/?""">U>?""">UEP?""">Ua?""">ri?""">9^y?""">_?""">?""">?""">?""">?""">r$?""">r԰?""">r?""">??   % TAALxassign_subj_obj_probs% F@4EMF++@ @$I@H<IDC>DC8DID@$$==_888% % V0//../% % $$AA( FEMF+*@$33B%BI'?""">7?""">G?""">qO?""">X?""">rh?""">Uv?""">r~?""">W?""">@?""">??   ( RpArialMonotype:Arial Regular:Vek%`h6`h`C`h4 hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% T&_AA& Ldcontract_pp% F@4EMF++@ @$IUE ?""">U?""">w-?""">H=?""">ZH?""">Q?""">Ua?""">l?""">|?""">8?""">?""">p?"""> ?""">??   ( RpArialsMonotype:Arial Bold:Versik%`h6`h`C`h4 hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% Tf*AA(Lhinsert_relpron% F@4EMF++@ @$IX#P !??% % $$AAFEMF+@H<NnC@@?@H<I""">>""">U5?"""> ?""">/?""">:?""">cD?""">U5T?""">d?""">gu?""">c?""">rL?""">r?""">r?""">?""">??   % TClQAAOLlinsert_subconjs% F@4EMF++@ @$I""">?""">a?""">3#?""">U3?""">>?""">G?""">aW?""">h?""">x?""">?""">?""">?""">?""">B?""">??   % T kAALlInsert_negation% F@4EMF++@ @$I""">>""">?""">?""">r%?""">90?""">:?""">I?""">Q?""">0c?""">l?""">?""">?""">k?""">rT?""">U=?""">9ƥ?""">??   % ToAALpinsert_infmarkers % F@4EMF++@ @$I""">>""">>""">9 ?""">?""">$?""">z-?""">rL=?""">rN?""">9Y?""">i?""">z?""">(?""">?""">r?""">?""">r?""">rd?""">r?""">??   % TuAALtInsert_prepositions% F@4EMF++@ @$IP !??% % $$AAFEMF+@<0Ne>@H<I>@% % $$AA( FEMF+*@$33B%BI""">r{>""">>""">q>""">U>""">>""">9.?""">U ?""">9N?"""> '?""">0?""">j@?""">QH?""">#X?""">Ug?""">9.v?""">|?""">3?""">r\?""">UE?""">9.?""">!?""">U?""">9?""">?""">ڰ?""">ø?""">?""">h?""">??   ( RpArialMonotype:Arial Regular:Vek%`h6`h`C`h4 hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% TAALassign verb position features% F@4EMF++@ @$I ?""">ǡ0?""">s@?""">ZH?""">Y?""">k?""">s?""">9ƃ?""">?""">?""">??   ( RpArialsMonotype:Arial Bold:Versik%`h6`h`C`h4 hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% T#bAA# Ldassign Case% F@4EMF++@ @$I8A18A#8L8Y868X48#818>8I8I8=I86I8X4I8>818#88868X48#818>8I8I8=eI864I8X4'I8>818#8'8486e8X4s8~#8~18~>8sI8eI8=I86I8X4I8>818#88868X48#818>8I8I8=AI86I8X4I8>818#8886A8X4N8Y#8Y18Y>8NI8AI8=I86}I8X4pI8e>8e18e#8p8}868X48#818>8I8I8=I86I8X4I8>818#88868X4)84#84184>8)I8I8=I86YI8X4KI8@>8@18@#8K8Y868X48#818>8I8I8=I86I8X4I8>818#88868X48#818>8I8I8=eI864I8X4'I8>818#8'8486e8X4r8}#8}18}>8rI8eI8=I86I8X4I8>818#88868X48#818>8I8I8=@I86I8X4I8>818#8886@8X4N8Y#8Y18Y>8NI8@I8=I86}I8X4pI8e>8e18e#8p8}868X48#818>8I8I8=I86I8X4I8>818#88868X4)84#84184>8)I8I8=I86YI8X4KI8@>8@18@#8K8Y868X48#818>8I8I8=I86I8X4I8>818#88868X48#818>8I8I8=eI864I8X4'I8>818#8'8486e8X4r8}#8}18}>8rI8eI8=I86I8X4I8>818#88868X48#818>8I8I8=@I86I8X4I8>818#8886@8X4N8Y#8Y18Y>8NI8@I8=I86I8X4I8|>8|18|#88868X48#818>8I8I8=<? % % $$AA( FEMF+*@$33B%BIW&?""">7?""">G?""">ZW?""">rlb?""">Uk?""">{?""">?""">ߌ?""">??   % T(]AA( L`insert_wie % F@4EMF++@ @$I""">?""">?""">r$?""">94?""">??""">I?""">X?""">rc?""">9s?""">G}?""">?""">?""">h?""">r\?""">UE?""">??   % TRi`AA^Llinsert_reflexive% FEMF++@ @<0Ne>@4(OCAܡCAܡC$A@$$==_888% % W(E (<(<% % $$AA( F\PEMF+@<0ϣC~սAܡC&BC~սAϣC~սA@$$==%  % V,?H!z|<|z|%  % $$AAFEMF+@<0Ne>@4(CV6BDV6BDqOB@$$==_888% % W(, 5  >% % $$AA( F\PEMF+@<0!DRJKBDYlBDRJKB!DRJKB@$$==%  % V,2< . ~ . .%  % $$AAF|EMF+@<0Ne>@, DrBD|B@$$==_888% % W$ U l d % % $$AA( F\PEMF+@<0!DBDBDB!DB@$$==%  % V,ir  ~  %  % $$AAF|EMF+@<0Ne>@, DDi CD< C@$$==_888% % W$     % % $$AA( F\PEMF+@<0!D'2CD)'CD'2C!D'2C@$$==%  % V, y ~ %  % $$AAFEMF+@<0Ne>@<0DBCD%PCxB%PCxBkVC@$$==_888% % W,i  *   g % % $$AA( F\PEMF+@<0BBaUCxB]CBaUCBBaUC@$$==%  % V,foW  fW W %  % $$AAFEMF+@<0Ne>@4(OC@, ܡC^CܡCٓC@$$==_888% % W$BE)<?<|% % $$AA( F\PEMF+@<0ϣCTCܡCP|CCTCϣCTC@$$==%  % V,?&H/zk<kzk%  % $$AAF|EMF+@<0Ne>@, ܡCBCܡCC@$$==_888% % W$BIE`<<% % $$AA( F\PEMF+@<0ϣClCܡC5CClCϣClC@$$==%  % V,?\Hfz<Sz%  % $$AAF|EMF+@<0Ne>@, ܡC' CܡC C@$$==_888% % W$BE<<B% % $$AA( F\PEMF+@<0ϣCCܡCCCCϣCC@$$==%  % V,?Hz1<1z1%  % $$AAF|EMF+@<0Ne>@, ܡC 8CܡCm!C@$$==_888% % W$BE<g<% % $$AA( F\PEMF+@<0ϣC}CܡCCC}CϣC}C@$$==%  % V,?Hz<z%  % $$AAF|EMF+@<0Ne>@, ܡCOCܡCD@$$==_888% % W$BE<< % % $$AA( F\PEMF+@<0ϣCbCܡCDCbCϣCbC@$$==%  % V,?Hz<| z%  % $$AAF|EMF+@<0Ne>@, ܡCDܡC D@$$==_888% % W$B!E8<-"<k#% % $$AA( F\PEMF+@<0ϣC#f DܡCyDC#f DϣC#f D@$$==%  % V,?5H>zZ#<#Z#zZ#%  % $$AAF|EMF+@<0Ne>@, ܡC?DܡC4D@$$==_888% % W$BWEn<%<&% % $$AA( F\PEMF+@<0ϣCDܡCDCDϣCD@$$==%  % V,?kHuz&<B'&z&%  % $$AAF|EMF+@<0Ne>@, ܡC#DܡC(D@$$==_888% % W$BE<(<1*% % $$AA( F\PEMF+@<0ϣC~(DܡCȑ*DC~(DϣC~(D@$$==%  % V,?Hz *<* *z *%  % $$AAF|EMF+@<0Ne>@, ܡCW1DܡCrL6D@$$==_888% % W$BE<V,<-% % $$AA( F\PEMF+@<0ϣC 6DܡC8DC 6DϣC 6D@$$==%  % V,?Hz-<.-z-%  % $$AAF|EMF+@<0Ne>@, ܡC>DܡCeCD@$$==_888% % W$BE</<0% % $$AA( F\PEMF+@<0ϣCCDܡCEDCCDϣCCD@$$==%  % V,?Hz0<k10z0%  % $$AAF|EMF+@<0Ne>@, ܡCoLDܡCWdQD@$$==_888% % W$B0EG<3<Z4% % $$AA( F\PEMF+@<0ϣC!QDܡC5SDC!QDϣC!QD@$$==%  % V,?DHMzI4<4I4zI4%  % $$AAF|EMF+@<0Ne>@, ܡCYDܡCI^D@$$==_888% % W$BfE}<6<7% % $$AA( F\PEMF+@<0ϣCѭ^DܡC`DCѭ^DϣCѭ^D@$$==%  % V,?zHz7<187z7%  % $$AAF|EMF+@<0Ne>@, ܡCgDܡC<|lD@$$==_888% % W$BE<9< ;% % $$AA( F\PEMF+@<0ϣC9lDܡCMnDC9lDϣC9lD@$$==%  % V,?Hz;<;;z;%  % $$AAF|EMF+@<0Ne>@, ܡC}uDܡC.zD@$$==_888% % W$BE<E=<>% % $$AA( F\PEMF+@<0ϣCyDܡCv{DCyDϣCyD@$$==%  % V,?Hzr><>r>zr>%  % $$AAFEMF+@<0Ne>@<0ܡCODܡCDxBDxB{D@$$==_888% % W,i E-<@<ZBZBB% % $$AA( F\PEMF+@<0BBRZDxB3dDBRZDBBRZD@$$==%  % V,f*o4B3CfBB%  % $$AAFEMF+@th CrCDrCD~CDCD[SCDZSCCZSCCCC~CCrC@C^( $$=='^C%  ;!6$!X(h$h$$Y$$X([[!=<>G % $$AAFEMF+@<0Ne>@CrCDrCD~CDCD[SCDZSCCZSCCCC~CCrCDrCErD~CErDCDZSC@$$==_888% % ;!6$!X(h$h$$Y$$X([[!$!X(##$<@H% % $$AA( FEMF+*@$33B%BCrC@0$9=ARIAL6@ Label.xml>""b>9[>""b>>""b>9.>""b>>""b>Uu>""b>C>""b>?""b>r?""b>??   ( RpArialMonotype:Arial Regular:Vek%`h46`h4`4C`4h hB`wTwwpww 8&2 w `$wwwww8&2PD38&2 -w w!dv% T* AA  L`Label.xml % FEMF++@ @<0Ne>@, uCe*C%Ce*C@$$==_888% % W$ftf% % $$AA( F\PEMF+@<0C%>CCe*CCCC%>C@( $$=='%  % V, f#%  % $$AAFEMF+@th CTCDTCDrCDCD?kCD?kCC?kCCCCrCCSC@C^( $$=='^C%  ;6$X(h$Oh$C$Y$$X([C[O=<>(GQ % $$AAFEMF+@<0Ne>@CTCDTCDrCDCD?kCD?kCC?kCCCCrCCSCDSCErDrCErDCD>kC@$$==_888% % ;6$X(h$Oh$C$Y$$X([C[O$X(#O#C$<@'HR% % $$AA( FEMF+*@$33B%BCTC@0$9=ARIAL6@|pdet.xmlX>""b>U>""b>䘫>""b>g>""b>U5>""b>>""b>U ?""b>??   % Tx4#BAA@L\det.xml % FEMF++@ @<0Ne>@, uCIBC%CIBC@$$==_888% % W$;>t% % $$AA( F\PEMF+@<0C VCCIBCC.CC VC@( $$=='%  % V,8A  %  % $$AAFEMF+@th 4C81CD81CDъCD)CD$CD$C4C$CC)CCъC4C81C@C^( $$=='^C%  ;6$X(h$h$$qY$$qqX(AA=<>^G % $$AAFEMF+@<0Ne>@4C81CD81CDъCD)CD$CD$C4C$CC)CCъC4C81CD81CErDъCErD)CD#C@$$==_888% % ;6$X(h$h$$qY$$qqX(AA$X(##$q<@]H% % $$AA( FEMF+*@$33B%B4C81C@0$9=ARIAL6@ auxiliary.xmlr=""b>UU>""b>r]>""b>@>""b>>""b>>""b>*>""b>9>""b>>""b>?""b>?""b>8?""b>-?""b>??   % Tj2xAAv Lhauxiliary.xml % FEMF++@ @<0Ne>@, C.ZCC.ZC@$$==_888% % W$qtv,[,% % $$AA( F\PEMF+@<0L'CmCOC.ZCL'CnFCL'CmC@( $$=='%  % V,nwn,n%  % $$AAFEMF+@th CICDICDCDoACD CDCCCCoACCCCIC@C^( $$=='^C%  ;J6$JX(h$h$ $Y$$X([ [J=<>G % $$AAFEMF+@<0Ne>@CICDICDCDoACD CDCCCCoACCCCICDICErDCErDoACDC@$$==_888% % ;J6$JX(h$h$ $Y$$X([ [J$JX(## $<@H% % $$AA( FEMF+*@$33B%BCIC@0$9=ARIAL6@xprep.xmlr5>""b>#u>""b>>""b>'>""b>>""b>>""b> >""b>7?""b>??   % T|'AAL\prep.xml % FEMF++@ @<0Ne>@, uCrC%CrC@$$==_888% % W$t% % $$AA( F\PEMF+@<0CӅCCrCCR^CCӅC@( $$=='%  % V,L%  % $$AAFEMF+@th CaCDaCDCDSYCDCDCCCCSYCCCCaC@C^( $$=='^C%  ;6$X(h$xh$l$7Y$$77X([l[x=<>G % $$AAFEMF+@<0Ne>@CaCDaCDCDSYCDCDCCCCSYCCCCaCDaCErDCErDSYCDC@$$==_888% % ;6$X(h$xh$l$7Y$$77X([l[x$X(#x#l$7<@H% % $$AA( FEMF+*@$33B%BCaC@0$9=ARIAL6@pdzu.xmlk>""b>C>""b>>""b>U>""b>'>""b>U?""b>??   % Tp"AALXzu.xml % FEMF++@ @<0Ne>@, uCC%CC@$$==_888% % W$t% % $$AA( F\PEMF+@<0CCCCC7vCCC@( $$=='%  % V,44%  % $$AAFEMF+@th Cs<DDs<DD?iDD8DDie DDie DCie DC8DC?iDCs<D@C^( $$=='^C%  ; 6$ X(h$ h$!$"Y$$""X([![  =<>G* % $$AAFEMF+@<0Ne>@Cs<DDs<DD?iDD8DDie DDie DCie DC8DC?iDCs<DDs<DErD?iDErD8DDhe D@$$==_888% % ; 6$ X(h$ h$!$"Y$$""X([![  $ X(# #!$"<@H+% % $$AA( FEMF+*@$33B%BCs<D@0$9=ARIAL6@|pneg.xml9H>""b> >""b>9>""b>Q>""b> >""b>Ǒ>""b>z?""b>??   % Tx %AAL\neg.xml % FEMF++@ @<0Ne>@, uCPD%CPD@$$==_888% % W$U!tU!% % $$AA( F\PEMF+@<0CZDCPDCGDCZD@( $$=='%  % V,!U!!!%  % $$AAFEMF+@th Ce DDe DD1DDDD[DD[DC[DCDC1DCe D@C^( $$=='^C%  ;s#6$s#X(h$>$h$2%$%Y$$%%X([2%[>$s#=<>7G` % $$AAFEMF+@<0Ne>@Ce DDe DD1DDDD[DD[DC[DCDC1DCe DDe DErD1DErDDDZD@$$==_888% % ;s#6$s#X(h$>$h$2%$%Y$$%%X([2%[>$s#$s#X(#>$#2%$%<@6Ha% % $$AA( FEMF+*@$33B%BCe D@0$9=ARIAL6@ subconj.xmlG=""b>>""b>9[>""b>>""b>r>""b>>""b>C>""b>>""b>?""b>?""b>U(?""b>??   % TC0QAAO Ldsubconj.xml % FEMF++@ @<0Ne>@, uCD%CD@$$==_888% % W$JM$t$% % $$AA( F\PEMF+@<0CDCDCDCD@( $$=='%  % V,GP$$u$$%  % $$AAFEMF+@th CXTDDXTDD$DDP"DDN}%DDN}%DCN}%DCP"DC$DCXTD@C^( $$=='^C%  ;&6$&X(h$'h$($`)Y$$`)`)X([(['&=<>mG % $$AAFEMF+@<0Ne>@CXTDDXTDD$DDP"DDN}%DDN}%DCN}%DCP"DC$DCXTDDXTDErD$DErDP"DDM}%D@$$==_888% % ;&6$&X(h$'h$($`)Y$$`)`)X([(['&$&X(#'#($`)<@lH% % $$AA( FEMF+*@$33B%BCXTD@0$9=ARIAL6@pdes.xmlk>""b>Uu>""b>>""b>U>""b>'>""b>U?""b>??   % Tpy"AALXes.xml % FEMF++@ @<0Ne>@, uCh D%Ch D@$$==_888% % W$(t(% % $$AA( F\PEMF+@<0Cr!DCh DC^DCr!D@( $$=='%  % V,}](('](%  % $$AAFEMF+@th 4CJ(DDJ(DD ,DDs/DD@ 3DD@ 3D4C@ 3DCs/DC ,D4CJ(D@C^( $$=='^C%  ;9*6$9*X(h$+h$+$,Y$$,,X(A+A+9*=<>G % $$AAFEMF+@<0Ne>@4CJ(DDJ(DD ,DDs/DD@ 3DD@ 3D4C@ 3DCs/DC ,D4CJ(DDJ(DErD ,DErDs/DD? 3D@$$==_888% % ;9*6$9*X(h$+h$+$,Y$$,,X(A+A+9*$9*X(#+#+$,<@H% % $$AA( F$EMF+*@$33B%B4CJ(D@0$9=ARIAL6@Subj_obj_spell=>=>r1>>cp>>UՄ>>x>>r>>>>c>>?>r<?>9&?>5?>1??   % T6AALhSubj_obj_spell% FEMF++@ *@$33B%B4CJ(D6@|pout.xml`>33>U>33>䘯>33>g>33>U5>33>>33>U ?33>??   % Tx#AAL\out.xml % FEMF++@ @<0Ne>@, C-DC-D@$$==_888% % W$v~+[~+% % $$AA( F\PEMF+@<0L'C.DOC-DL'C,DL'C.D@( $$=='%  % V,+~+;++%  % $$AAFEMF+@th 4ClDDlDDoDD/sDDvDDvD4CvDC/sDCoD4ClD@C^( $$=='^C%  ;';6$';X(h$;h$<$=Y$$==X(AG % $$AAFEMF+@<0Ne>@4ClDDlDDoDD/sDDvDDvD4CvDC/sDCoD4ClDDlDErDoDErD/sDDvD@$$==_888% % ;';6$';X(h$;h$<$=Y$$==X(A""b>qp>""b>rܗ>""b>9N>""b>>""b>>""b>?""b>?""b>??   % T|(AAL\case.xml % FEMF++@ @<0Ne>@, CqDCqD@$$==_888% % W$vm<[m<% % $$AA( F\PEMF+@<0L'CarDOCqDL'CpDL'CarD@( $$=='%  % V,<m<*<<%  % $$AAFEMF+@th 4C'zDD'zDDT}DDDDw(DDw(D4Cw(DCDCT}D4C'zD@C^( $$=='^C%  ;>6$>X(h$V?h$J@$AY$$AAX(AJ@AV?>=<>G % $$AAFEMF+@<0Ne>@4C'zDD'zDDT}DDDDw(DDw(D4Cw(DCDCT}D4C'zDD'zDErDT}DErDDDv(D@$$==_888% % ;>6$>X(h$V?h$J@$AY$$AAX(AJ@AV?>$>X(#V?#J@$A<@H% % $$AA( F$EMF+*@$33B%B4C'zD@0$9=ARIAL6@Verb_position.81=>c=>x7>>r\]>>Q>>U>>>>r<>>?>( ?>?>a?>3)?>U9?>??   % T5AALhVerb_position.% FEMF++@ *@$33B%B4C'zD6@THxmlß>33>U5>33>>33>??   % T`  AA LTxml % FEMF++@ @<0Ne>@, Cs<DCs<D@$$==_888% % W$v?[?% % $$AA( F\PEMF+@<0L'C*#DOCs<DL'C2~DL'C*#D@( $$=='%  % V,@??@%  % $$AALdi3h2)??" FEMF+@ CompObj>qObjInfo@VisioDocumentGVisioInformation"A ՜.+,D՜.+,X@HP\p |   Microsoft PagesMastersPage-1ProcessDynamic connector Direct dataDynamic connector.88Visio (TM) Drawing GTRh !fffMMM333^C8 TZ Arial@:NdZWingdzs@vm Monotype Sort NuSymbol5T?? Y@-1U J:DT1EW-hTT<* /Ub bO0zGz?@8@H2!kWb*Uk9 +HPL/^&9^$?Ak^&*,,'%/v&&U  1y   )? 2    12?aBBHEHEHEHEHEHEH@?>?:`T2BBHEHEHEUHEHEHEHE%H@%O9 F7AOY@;P AsVsVA!gLTkY W_W__ !`#k4lb6u`kW *4l 4l %Y?P:?-\ #!+|QtKf2|2|2|I2wGQAUoTMeE$ttA%_8BOTOfOxOO??HO?7ܻXuW?YsU42 T*?P?޾B~;$^ ????+P?OO+O=OOOaLqOOOOOOOO__%_7_]?Gh_Ռ249_ﳸ____ oQ%+o}RodovooooooooI_[VK}??? 1ASewяR#ЦUf/ ҟJ/\/n/1CUgyӯ -?Qcuڿ$\yԁ1λ|Z|(g${iPхϠA32O?贁Nk@RFRR#Y^Ç?(:iӿٟ` S odXXLetterO_b PRIV@|rpE fg4o/#/5/G/Y/k/}//////// ??1?C;} Display.10UFDfP h-RTUUUA@ ?I? 3h  ePqYk  KHo:Lw >2QO` Flowchartw`u D(2OVhz~ WThis symbol represents any kind of processing function. Double-click to add sub-page.H D  # =hj0>TdYY9 ͉UA@ ??P6 u` 6u e.A&a8 e&0LgH>v5 LS{5 `7Copyright 1999 Visio Corporation. All s reserved.`_BFS.chm!#22268D" 0d9 l>YUdv "}T <b;2F6& :2ge24 7=H?{1>%??K91{1^;.6=G=E%2O;rA wB}C! ai;(!aOj_(O_k,OqV(@^0A^ 5F@dL?VVG~#A&QR?_\5C _T oo!TS>%[fojoQ P3m5#_d^*+T$]tjt>"e}sIsQ q?B`Cost6,Enter the cp associated withpis process S"`@ zq`vQ:bsDu* wd@ {ofsstep @zq>%$s Resourp:wnumbpdpeop{le quirpktopm܀testask n>!auY=@|v=# %Properti:0Se}tsustomqsce sele}cqshape s{ wBb  f ,f2^*mb?ؿ贁N8i6??54 _ Q);M_>m w =_H'-D !OyaGE_5>F\Rw#C MB L!c]faGɬ]@k]P+84aT|p8= IUFDf h-TUU[U@@??I?`d buoqYkQhu23u` Connector ` e1Crw UH ^   -}|5 p`I`V?}`xa833ސ]3σ3u 33?, ?Gxpx^& CThis connector automatically routes between the shapi ts.b?2jZ/0? HD @# =h8T YY9 BT#F oU@? P6 u `u bA@]u  .(#DB uu`h?\hr|uVa@-?bl;'bE-ho'$y( 2rq?@I ?$%? @"U*5L -br  ^vv"(2uI."q28v"uh9Bd&</MS{ #145 `Vis_BFS.chm!#22291`7Copyright 1999 @io Corporation. All $Ud vE \/4 *&1$b24R(^[w[D ZQi a59 93O'2"q?/g;2GHu-!N !OyaGE_L>Fd:Kw#@7B T?P$do@+k"sAkoUFDfP h-RTUUUA@ ?I? 3h EePqYk Ho1C*w Q2Q` Flowchartn`   @?o2 MN|_X_ r pN u|jbfUM^Data that is directly accessible, such astored onsk drives. 2mb?ؿ贁Ni76??k4 UH LED D# Uh4J T]]TMANMUAZ@ ??P m>uA` uVM J.AJ<-QJ*(.88YFxdnJ>ʷI@>brM J?V,"rZ|-#%))<"<>h>@/<@SNG1L4+Ir"`?Copyright (c) 2001 Microsoft Corporation. All q2s reserved.l`Vis_SFB.chm!#265119S3!tA7 30lJ0JUhh5!*q+#`#5brn"u$b@$ `UD"B($A'3`#N89 _EA$^EA85# OG>:%?RjK_ lP\RV:#{^( bPA,b(EGOO@a`Y?ffx3MIaofBDa(2v#b%d@oA}a5Yi'Z yI@Oi85k {$_gAQA(D<5oicOUwKHa^)y@Q%8_[?n[RVdMle85ki\f"\#B2#FA@CA -@B-J;#mQhe$67+B>T11bR"o2sN G5G1a b`Cost6f3,Enter the c associated withѠis procgess0 B`) %`a#Du3*f3 ̧dE {ofstep0D5{$( Resouri0:̧numbϠipeop{le0quirktoՠmte/task0nZ:LJA7O1G1I3 %Pr?operti0Rb ł01Bce>՟I5z{~!g)dP(:hTA?"aa2(QG1e`Dct,data,dLly,acible,such,sto,onVsk,drives,Basic,Flowchart,inoform2,f~Vagram,joiners,ۥ0!5S $@=H-H? !OyaGE=>FQ!#dTKB tURa,Bo@+*ko5aDV*PUFDf h-TUUU?@ ?6I?d XboqYk*Qu23u` Connector} ` e1Cw \UH ^   -}|5 p`I`V?}`xa833ސ]3σ3u 33?, ?Gxpx^& _Connector that automatically routes betweene shapit cs, using a right-angled line.b?2jZ/0? HD # =hj8>TA YY9 T#FAoU@? P6 u `u bA@]u  .(#DA@uu `h?\hr|uVa>b@-?bDl;'IbE-ho'y( 2rq?@I ?$v%? @j"*5LA-brB  ^vv"m(2u."q28v&"uh9d&<?/M) #145 `Vis_SFB.chm!#26514I`?Copyright (c) 2001 Microsoft Corporation. All $>U@dv5 !\^4 *L1$bC24R(f[[D ZQiAa59 93O'2"q?7g;2GB81aj e/}@~!gdao7 T KaKa 2(c1B1@Dynamic,connector,autom[@cally,routes,betweenus,Fv%w#{J&7B D}'gdo@K+k"s~ (PUt4> ףp=@bq?#@ u,}C-$7"A8t4> ףp=@bq?#@ ,A>-7U=?@t4> ףp=@bq?#@ -}C-67"A?@It4> ףp=@bq?#@ -A-7; @ ,IR@,BR@8-IR@-KRH<( H<( H<( H<(  E. RE4. RE. RE$. R_6(T ݦS /- BV!9/{"N$/"s+/?\9-PD~,. AM1AM1PPTPPT1h/:0 A/ԏG0 A/T0 A/a0AV7V? ?{ @`@g "*^p? N(m*~441CUFP (?? @`*7M uDTey tahm$/T tqUI> ףp=@Ibq?#@? ?I.CUg`tvq 02U468:E<>*B$e2-?&-(qu` Y"c"i"u$s) 2.%?Ad-'/sU 16t*?ot!%Q11 ` Flowchart `2 :1@??9ConnectoIr?!Je*&(2 a*\4 .AT0 JY!1Ri!TAUUQ4!"#%&')*+,-J./1357$9;=.AH~Q@TW%TWTaWTAWAXAWAX_1WAXAW_5TAWAX.AW.AXAWAXWTAWAXAWAXY!WAXAWY!XAWAXAWAX1W1XAWAXWXi!Wi!XAWAXW^BWAWAXAWAXAWAXAWAXAWAXAWAXAWAXt!Wt%TQWQXQWQX QW QXWXQWQXQWQXQWQXWT"QW"QX&QW&QXW*QXBQW.QXW2QXJQW6QXW:QXRQWXWBQXZQWXWJQXbQWXWRQXjQWXWZQXrQWXWbQXWX*WjQX!W@XWrQXCZ@XeW5TEVuP*AARQRQt!t!QQaQQA Q QAAQQ_1QQAQQ.AA"Q"Q&Q&QA*Q*QA.Q.QA2Q2QB6QD6QA:Q:QAD1BQBQADi!H1H1AybQbQA}A#'AAWAAXNAY[AcZht![uQ\Q1] Q^F_Q0QaQcbc"Q&QrQrQ..BQ1JKJQT+U8RQVE#dRZQKAKAƻflbQgyjQ1ijrQklm*cn!p@q@z1r@es@ou-?Q?eT@11D_5AX4:UA0ey"I4Q!U!|Gz- Ӯ#i@8%5..?J*(RQ!wLi!B  (Ay#BS1A4^һ00XAQ3#@h QPQ vFo3AR5ڱX0cLX? L1AAX0U0. Pre-Processing+ OGX8A mFċF<5^4m?A_3UF?:e?VQ[DpRuUE?X?W?@Ae% R?bOtLAB@ALKOT8O=D_A&_“@<(I 5t)(7$)K'2qPMa1:oc'%aTIIe"VVcc}ak}/ > ףp] @-+4fX]PbSv(e"]Fїv Ҭ//*/PbtφϘYfxd dDVhz!ߓߥ߷ߌq6 1s,>PbtGYk}!ʼn+$5create_lexnodeBT?xJ_,>obtR_0o#o(oo9,8*E*"=Ͽ);./@/R/d/v/*1/$zE////?B_%?7?I?z"b?t????7w 6? OO1OCOUOgOyOOOťOuDZ&DeepDegraphLF!_3_aW_i_{_)o_ӿ____ ooQAoSoeooooo:C(|(sߎR 1CUgy)e㏪$i(:Lp|Gz?@]MW[@ʟܟ8h3hA@Sew/үT(iI[I. Flesh-Outj|Ŀr/I/:/^ϑ/ϔϦ1d=?qq ?Ӕ&;{O? ߒ?QrX ls`ODVhz ӉA,AI_m> ףp] @,>AsfIQvsfA7IQ ,a@TUmUAB1insert_determiners_ //0/i f/0//////2?,?;OP?m,ߑPПְٟo79O"po*宯OOOOOOO __._@_R_Qt_>U_____!oo%o>oPobo%No`O66 g1c fooo 1CUyCUPbt߆ߘߪ߼b x%O!%[!vIsauxiliariez/pS_ .@Rd___7*{o ooA?I"o$ 48K8y //-/?/ \/ѷ|/ۏ/// _// ?7)?;?M?_?q?"ɦՖIA;Ց4IA????OO)O;OMO_OqOƅۯO܅IՑOiCexpletive_s?ubjectOO* _2_D__h_______& oo.oqRoOϹoooooYAwN>rbJ՟] 1CUgyO)Tq$Tʏ0]6HZ1L w41$>ҳ3> /ASewm01IFassign7_oprob?QuGϽ);w/_qσϠ/ !?%߮`?i?w>?ߏ?߳?t33.@Rdv!3EWihё66%6f2a`rONk QTlUCU M!contract_pp_//A3/E/W/i/{///// / ?O/?ES?iw?O6Fo?֟O9hSm_w_$__}/7o_<"o0/o  1CUt1sO//$/0jZ/S%I/[/m//2+)f*Rf#/// ??/?A?S?e1ONkp?2D?Tu?uq8assign verb position feature?O?5OGOYO_}OOOOOOO?_1_C_g_h_____no(1-gcoWχo{_eI[mj+zOas:#A/ԏ!M$'g#&%7I[mX?j?|?= ?3=I3negaLYk,aׯ 1CUyiӿ?//ol@/[/8Kc/ooo"4FDVhzߌ߲yHH'O T7I[mxqc!V'q#5GYk}ٝE'o(EQ!a1i)infmarkerXHvOl~>o 2jVhzo////S\/=Y,>X//2%]#2?#?5?G?Y?k?}???????BOt:OLO^OoOOOmOOOO _ (3n W1(IV\_n__________t:ouB l preposi3ooïooo;%7I[mߣRd3Eci(F'r#^pʟܟ$9CGcϯv5_"*3D+;@ׂ3DA"G  UC_@(S[Uٖş؟h_zRY*_U_____ o2oDoVohm21#)ooolhotᵿؽCUgyW| %:L^p ʏr+"3'MEy ,uA #*GkЂm!GZ11C(S[Upsb"yRDi{[ïկYrc%x#J\nt?ĿF]$"i )1C?n@r);M_q߃!ߤ߶ B m GkD//)R S6/?sb@U7%7I?m?]\d2NF@)Dt@Os]@);AQa_ `r> ףpb ppe W/i/p i0/k`|Gz?@?]M})_4v?3Eӓbevr>?%Q.|v^4wo?i????}?OO$O{UVuOOOGbaOQOQ,X_j_UώXʙi_@(`OQ_Ǹ_oo)o;opMo_ob@"s$@-@-/oi5&o~%b=Wb] (Z u.›);~|h{ryXT@%5GYkRߏŏhPեԦ$6]5fVX{B)T:S١tdǒ 9.V.ٯ!3EWiW/$/6/H/o4'%#15'&P2Y9xρ? /AOPObOtO OOOOOO_ q6ŵ_ƟSe_CX_oo(o:oLoݯpo@oooۯ`_gQo Cw80GYCvC?@?R??,>֣IѦdЏ '9K]o 054FS4sAS֡\r̯]@&[J$i#x÷(ǿ$Vc V.dϝψ>'7Cu,>PbtEv5_֞߰+?-?I"nה//R$dS6/?RCYX7(:L?p@?d`_g2cQF C8x$GYtCOv]CϮ@R,>AGTAd_+/'/9/K/]/o//".'QD5v?4F1vsA?SE`wr??????]?OO&O[JUVriцxOOOGqOٴMcOM._d____®!_T!_uoo,o>oPoboto 0jZ/oooooo('qbDƿ!X*6h/8B}(:Lp`,QQ C80GY!v!ҟ,>*DVdЯ:Є@@`մ^|> ؖ' '9^u3`u k^p?@|Gz?@iWz?A $0%76$Alb2₌trϪ 'ЖϨϺ-&5'>(`UC_ yߤG׸HD)/L-,)3) U[m _.@/R 1G/Z0L3_1+;9c1?l"^22/3)e1;e+*uQ@@]MW@?@:E '#1憏FBSe0+ JI@3:H3B@NO>q˂-^C~H2Fbb^bCzqM#J mb/2T_'8+BTCIT"TTd(Լ92G%?88P[omle3.?GUC1//Nhcu.` 暀HCG9Fc1>/;Label.xmlƑ1Ƒ? %Ha2|h6aE`1l IF(N)׳ URI>(H!B(GY4Lϖ!p4hqƑ (׭OewA!3TGş_1 #MK 6&J4!#O0O贁NkA#%&`,_ȿ_Rd___o(o:oLopgߌe/ooboG?,'?m/?x{auxiliaryvwO 2DVhz\՞nոʏ ܏^.+n =GrVf_ѦgΟ'9KbaavЯdoi*<Nbal}baڿ/O"4 0B6GE*(o k)<Vhzďя .@RdvϾ/*<rɐ[mVh߈,>,/P/b/t/P///O'C%`?:??^?x??:'??4O%Hm4HprepHFzUOzrOOOOOOO_ (QV䐢c/YcA_S_e_w__>(VRVWQ8&_7/obzi'aDoVo\=Oooo!noe(3'9Ko!@OcM!6HZP~tϚߑߣD4_Xj|SϠğ֛ -!9&fb[/toٯ!3E+_=_{ÿտ υ/A_ewωxR߿e+=asT_f_m\m$OOOl9Ko 2VorѿH& .nR϶Ϛ  zu^c+=OasCeUfC"'U`,A.'Yi!M/r/"뱍)o//?on? ?2?`H>m]?{!????@>??OO*OT`SO `yO˯OO6 _&r_:__%/___ďo#o5oGkL`Xojo|ooo;a#6"2oo&/J\n"4FXj|On/֏ß0Tx_ҟAONO>Poom ݯ___7I[0Ƿ#?@u?!ϹEW{ύϟݏfp% ^neg(Z5TZRӟv߈߬߾&!3EWi{ AeZhA?$6<?/>}б&\nбBб(:0^T_xbotooonԥ/8/J/\/3_///+//// ?̱=Ѷ;?_?q???S????OO%O [OmOOOOOOOOd_!_E_W_i_X2o_E___ ooAoSo3Eofa;?o^=aj|ߎL/*/N/?d6ROl(Ə_M_2?__zъsubconjA=$AŸGo .@Rd4ǯٯ4 85Hv,QcAlwύD'81dgס &ep7I[mC«ǿ);M_q~%7I[R?//_3/?W/i/{//n//%2ߧ?סN?סt?ޟ??O(O:O^OpOO_GE$OY__$_6_yZ_l_~__Jh /Y[es fo䫑6o/Zolo~ooooooo);M_|epq|Awq6?ryKG% asסԏ O_@Rdvo%ja̟ B8\nFXj|Ȭv1ۯQdv1x.@dv1֟L֋ ÿտOBvuqOCU1nv3ϬϾooobt?Oυ:L^p[oo??11Ͽ1W///yϯ ASe߷OF_{f=Oaﮯ7<QcaseQtQ#GYk }ِ F%K);9QF^E oƍϲQ :_ /O^N`rQ1] "/4/F/X/j/TT/Q/ ?/%?vKX?0BTf?1?@1?g OO-OQOcOuOE4jZ/OOOOOѯ/ed _`0_B_T_f_%_______,o>oPobotooooo6?oo(:)p\?$//~R _0]RՏRodvψϚϬ Ucb#5ҳ>fx~_q߳R܉9K]oTk*X_j_|_r_?ŌOOOOK !"#$%&'()*+,-./012589:=?@IJKTUVWXYZ[\]^_`abcdefghijklmnopqrsy*}t4F> ףp=@bq?#@ 4sC-ܔg A@ēmtLRH<( E4t R\,}1Ans@?4t.PDdt.PU1( UO"D&aU=QJf )h"TyU-x +3 5 7 9 ; = ? A B C E Ʌ&Q- H*9(T#--rQ//,GuideTheDocPage-1Gesture FormatVisio 90ConnectorVisio 00Visio 01Visio 02Visio 03Visio 10Visio 11Visio 12Visio 13Visio 20Visio 21Visio 22Visio 23Visio 50Visio 51Visio 52Visio 53Visio 70Visio 80Flow NormalHairlineFlow connector textProcessCostDurationResourcesProcess.3Process.4Process.5Process.6Process.7Process.8Process.9Process.10Process.11Process.12Process.13Process.14Process.15Process.16Process.17Process.18Process.19Process.20Process.21Process.2Dynamic conn?ectorDynamic connector.25Dynamic connector.28Border GraduatedDirect datavisKeywordsvisVersionBorderDynamic connector.8Direct data.50Dynamic connector.51!Border Text Transparent LeftDirect data.52Dynamic connector.53Direct data.54Dynamic connector.29Dynamic connector.30Dynamic connector.31Dynamic connector.32Dynamic connector.33Dynamic connector.34Dynamic connector.35Dynamic connector.36Dynamic connector.37Dynamic connector.38Dynamic connector.39Dynamic connector.40Dynamic connector.41Dynamic connector.55Direct data.56Dynamic connector.57Direct data.58Dynamic connector.59Direct data.60Dynamic connector.61Direct data.62Dynamic connector.63Direct data.64Dynamic connector.65Dynamic connector.23Direct data.66Dynamic connector.67Direct data.68Dynamic connector.69Dynamic connector.24Dynamic connector.26Dynamic connector.27Dynamic connector.42 3Iyv E3Ov E3_v E3\uvG3tuv E3uvE3uv E3uv E3uv E3uv E3vw E3v w E34vw E3Lv'w E3dv4w E3|vAw E3vNw E3v[w E3vhw E3vuw E3vw E3 ww E3$ww E3yA3lByE3PyA3TyA3XyG3syG3ԉyG3yG3yA3yE34yA3<yE3TyE3lzA3tzA3| z E3zA3zA3zA3!zA3%zA3)zG3ԊFzG3^zG3 }z(G34zA3<zA3DzA3LzA3TzA3\zA3dzA3lzG3zG3zG3{G3܋/{G3N{G3m{G3<{G3\{G3|{G3{G3|G3܌'|G3F|G3e|G3<|G3\|G3||G3|G3|G3̍}G30}G3H}G3$g}G3<}G3\}G3t}G3}G3}G3̎ ~G3+~G3C~G3$b~A3,f~A34j~A3<n~A3Dr~A3Lv~G3l~A3t~A3|~A3~G3~A3~G3̏~G  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~UnU U UUU;CLt4> ףp=@bq?#@ 8X}C-7A%t4 @5_ A-?57AJ@KR@A56RH<( H<( JEM RETZ R{  g"4FX (h(#v+@(u/`,Xz [La:YB_ P  bv&5I!`~2}F!R34s| /R$C0wP.,oxNO)MlU1'}"uh~\I8=DPS]<2 6 T?#*PD~g6%T`<d Cg k @SummaryInformation(DocumentSummaryInformation8B_1081078247 FWGPGOle K      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\o`abcdefghijklmnopqrstuvwxyz{|}~Oh+'0@HXdp|mgamonGn  EMF l@VISIODrawingLlno ??d((o[[[PPP@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((hhh`````````pqx}@@@@@@@@@@@@@@@hhhFFF@@@@@@@@@@@@@@@?XC^C^C^C^C^C^C^C^C^C^C^C^C^E ߿߿߿߿߿ؿ߿߿߿߿߯@@@$EC^C^C^C^C^C^=C^#2C^C^C^C^A[$ENNNOOO&PC^C^C^C^C^*;C^=C^C^C^C^C^;R #.k 888tttOOO (V6LC^!/yC^!/yC^!/yC^!/yC^!/yC^!/y;R #,d```翿翿翿翿翿翿翿翿翿翿@@@6C^C^C^C^C^C^C^!/y6@@@@@@@@@@@@@@@:::nnn@@@@@@@@@@@@@@@=UC^C^C^C^C^C^C^C^C^C^C^C^C^LmmmCCC}}``````46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?@@@@@@@@@@@@@@@hhhAAA@@@@@@@@@@@@@@@0C2F2F2F2F2F2F2F2F2F2F2F2F2F .߿߿߿߿߿ѿ߿߿߿߿߯@@@">C^C^C^C^C^C^C^C^C^C^C^C^C^A[6gggOOO&PC^C^C^C^!/yC^!/yC^!/yC^!/yC^C^;R #.k666gggOOO&PC^C^C^ C^C^C^C^C^C^;R #.k```翿翿翿翿翿翿翿翿翿翿@@@$EC^C^C^C^C^C^C^C^C^C^C^C^C^A[$E@@@@@@@@@@@@@@@>>>rrr@@@@@@@@@@@@@@@?XC^C^C^C^C^C^C^C^C^C^C^C^C^E CCCpqx}``````@@@@@@@@@@@@@@@cccAAA@@@@@@@@@@@@@@@߿߿߿߿߿ѿ߿߿߿߿߯@@@@@@```翿翿翿翿翿翿翿翿翿翿@@@@@@@@@@@@@@@@@@>>>rrr@@@@@@@@@@@@@@@LLL``````ϿwwwHHH@@@@@@@@@```````````````QQQ```````````````PPP``````ϿwwwHHH@@@ ppp@@@```````````````QQQ````````````PPPppp:::``````߿wwwHHH@@@ppp@@@```````````````WWW````````````PPPyyyIII``````߿wwwHHH' ,r!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y &'@@@.C^C^C^C^!/yC^!/yC^!/yC^;RC^C^C^¿}}}!)UC^C^C^2FC^C^C^C^C^C^=U "+\FFF@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((#KC^C^C^C^C^C^C^C^C^C^C^C^C^;R "/r@@@ 'MC^C^C^C^C^C^C^!V 'M```````````````ZZZ````````````PPP2FC^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^ &PPPEGO6">">">">">">">">">">">">"> LNW```￿```߿߿߿߿߿TTT߿߿߿߿hip"T!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y &%&/@@@@@@@@@@@@@@@www@@@@@@@@@@@@@@@@@@=UC^C^C^C^C^C^C^C^C^C^C^C^C^ C^C^C^C^6LC^2FC^2FC^A[C^C^=U "+\[[[((( C^C^C^C^C^C^C^C^C^C^8O &"/r@@@ C^C^C^C^6LC^2FC^2FC^A[C^C^?X!)U @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@xxx'8OC^C^C^C^C^C^C^C^C^C^C^C^C^ & 翿翿翿翿sssGGG翿翿翿翿翿'">">">">">">">">">">">">"> EGO``````߿߿߿߿߿TTT߿߿߿߿hip"T!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y &%&/@@@@@@@@@@@@@@@rrr@@@@@@@@@@@@@@@HHH=UC^C^C^C^C^C^C^C^C^C^C^C^C^ ߟC^A[C^2FC^2FC^2FC^2FC^2FC^=U "+\```߆333 C^C^C^C^C^C^C^!/y8O &"/r@@@ C^A[C^2FC^2FC^2FC^2FC^2FC^?X!)U @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@hhh;RC^C^C^C^C^C^C^C^C^C^C^C^C^  翿翿翿翿sssGGG翿翿翿翿翿|}'">">">">">">">">">">">">"> <>G``````߿߿߿߿߿TTT߿߿߿߿rsxL!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y .,.7@@@@@@@@@@@@@@@rrr@@@@@@@@@@@@@@@HHH=UC^C^C^C^C^C^C^C^C^C^C^C^C^ ߟC^C^C^C^C^C^C^C^C^C^C^C^C^=U !)U```߁*** C^C^C^*;C^C^C^C^C^C^;R "/r@@@ C^C^C^C^!/yC^!/yC^!/yC^*;C^C^?X!)U @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@hhh;RC^C^C^C^C^C^C^C^C^C^C^C^C^  翿翿翿翿sssbbb翿翿翿翿翿|}E!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y546?NNN```߿߿߿߿߿TTT߿߿߿߿|}'">">">">">">">">">">">">"> 67?@@@@@@@@@@@@@@@hhh@@@@@@@@@@@@@@@HHH;RC^C^C^C^C^C^C^C^C^C^C^C^C^  ߟC^C^C^C^C^C^C^C^C^C^C^C^C^=U !)U߆333C^C^C^C^C^C^C^C^C^;R !/y@@@C^C^C^.AC^!/yC^!/yC^!/yC^C^C^=U !)U @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@hhh=UC^C^C^C^C^C^C^C^C^C^C^C^C^ 翿翿翿翿瑑bbb翿翿翿翿翿rsxL!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y .,.7NNN```߿߿߿߿߿TTT߿߿߿߿|}'">">">">">">">">">">">">"> <>G@@@@@@@@@@@@@@@|||hhh@@@@@@@@@@@@@@@PPP;RC^C^C^C^C^C^C^C^C^C^C^C^C^  ߟ C^C^C^C^C^C^C^C^C^C^C^C^C^?X!)U߁"""(((@@@@@@@@@@@@@@@@@@@@@@@@@@@888 C^C^C^#[C^C^C^C^C^C^;R "/r@@@ﴴ C^C^C^C^!/yC^!/yC^!/yC^!/yC^C^=U !)U@@@=UC^C^C^C^C^C^C^C^C^C^C^C^C^ aaaVVVhip"T!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y &%&/NNN```￿```EGO6">">">">">">">">">">">">"> EGOcccKKKPPPC^C^C^C^C^C^C^C^C^C^C^C^C^C^ & ߟ 'MC^C^C^C^C^C^C^C^C^C^C^C^C^?X!)Uccc@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@&&&$QC^C^C^C^C^C^C^#[8O &"/r@@@¿}}}!)UC^2FC^!/yC^!/yC^!/yC^!/yC^!/yC^=U "+\@@@.C^C^C^C^C^C^C^C^C^C^C^C^C^C^dddYYY' ,r!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y &'NNN``````'">">">">">">">">">">">">"> LNW___GGG```%&/6LC^C^C^C^C^C^C^C^C^C^C^C^C^ .ppp C^C^C^C^C^C^C^C^C^C^C^C^C^?X 'M@@@@@@@@@@@@@@@@@@@@@@@@@@@﨨HHH(((@@@@@@@@@@@@@@@@@@@@@@@@@@@888 C^C^C^C^C^=C^=C^=C^C^C^;R "/r@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@߮ C^C^C^C^=C^=C^=C^0CC^C^=U "+\@@@ppp ?XC^C^C^C^C^C^C^C^C^C^C^C^C^dddqqqabh ,r2F2F2F2F2F2F2F2F2F2F2F2F2FE'rrrDDD<>G46?46?46?46?46?46?46?46?46?46?46?46?46?46?``````WX_```,.74IC^C^C^C^C^C^C^C^C^C^C^C^C^5@@@ C^C^C^C^C^C^C^C^C^C^C^C^C^?X 'M@@@@@@@@@@@@@@@@@@@@@444MMM***@@@@@@@@@@@@@@@@@@@@@@@@@@@888 C^C^=UC^=C^=C^=C^=C^C^;R "/r@@@@@@@@@@@@@@@@@@@@@@@@ϛC^C^C^=C^=C^=C^=C^*;C^;R #,d@@@```@@@A[C^C^C^C^C^C^C^C^C^C^C^C^C^. sssUUUWX_!/y2F2F2F2F2F2F2F2F2F2F2F2F2F=@@@cccDDD67?46?46?46?46?46?46?46?46?46?46?46?46?46?46?@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ !!!@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@HHHǿ000ǿǿǿǿǿ>>>ǿǿǿǿ XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@RRR@@@XXX@@@XXX@@@XXX@@@XXX@@@000XXX ```@@@888ppp@@@ppp@@@ppp@@@ppp@@@ppp<<>>@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@翿:::@@@@@@߿000(((888@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@***:::@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ ``` ``` ```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((888@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@>>>@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@```@@@߿???"""```"""ǿ```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@>>>666@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@CCC@@@ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@_VPID_PREVIEWS_VPID_ALTERNATENAMES_PID_LINKBASE A  FMicrosoft Visio DrawingVISIO 6.0 ShapesVisio.Drawing.69q rٗotMa)L50>0v JZdU?ɵ~{hs^d1}'!e_W\յ~+$wӫ3ݗI`OYIfjI f]ɻY'ʵ$ 0O`+:=|bO'9<&D$}(?Ie})L09[WIENDB`}DyK _Ref530885112{DyK  _Ref11560948}DyK _Ref530885349 Dd~ p  C ,AbasictreeC"b "VݢQ n "VݢQPNG  IHDR"sRGBPLTEg&T JIDATxK0 @pU!W8MIH–eyΊG62"UB!yD Q<$?7bR$36m=I!&`RPq#KDѣ|CD{<xEG"H ďeD0!uau#2#hۦH"*R"Bv/ bVD&MfOHLkH*g ߏE{}){k gNmSe-?;^~=w_(2ȍΠ"!}+$Dn#$nVG lx$6lg%Tq1?(3!D~yE#2dd$XCTp" OzoLN? x8U=)#TL:ڣ$yAl?2X#>n;-""DQH>ډHYE"Є\Q*FGۑ#DWA+zbs:sIq[ǃԖ|sDpXUDdzxs-&w5j%A >DD9Cs|bN2(Q+їW9rk(Zz Թydg{H ȒD''dqŘ_j«9!f>\ܗW2 2~Pr+Ċ#S-W}FomD3DBA6HFYxG@@'y[GFJZEd@"c  5N 24!+}g"d$ý!s}D:7vd6YfN&~9dgBJpmN/rLU2}Ǜsίt*nK>+ki$L#A5&˛j?D..@wK:9LᜯYD[!#c<+^wr'"/V  چrdDNg-PxzhY-*%➱&sGa|ems ";ƈD/&$#94'ɹ߮mϋtۆmHN&¤3V*Ǚ(|Bhݽ.ڨNMёUHZs=}_{Ǐ1"R +S2wGc~UNhȇDfPZBD)ِHę(\kBudWGŘY6߀X[H$dG2gBu;>V1&8CNt|IA"6fҼ&VfR(bkt@i"%Md)IZSrG+FbHv~:Ad:,sX)Kqd>D,"E$N՞(%? m+RDH@BR.fXrbKU91* }ӜY3X"^}D;kDT~qBj5\T`g2$| 3 ̄ t~qBj^UZ^I,.~:nV1(\c_m4$\۫. ,/\+F¢5Sg63e`*]+xװ TfOդ}.r݃EU-;("d,~$Dzamd:]Y^kvLU٫#3>"V ="`__{%`Eu mq-9G?~{8o'ō9-C}X[$cctkw ~D}XWy#D&,zXsw$˴@~Y_Ia'&ڔ[ ,Dؤg+,w_VrCz ,'U:P.| G?%H¶".Bׅ>$]PcF{jXaET$`m%IXQ;J+c*0t(~e_<ܩ3JWn+5Vbn&6C6D[e^Iojz { frkA{ ߞgϪ7ڗ&-WQ ֪ocHfm#Ӟ#g^ܱq[ Y{vٳUa}jU%oۣCLgm"k*` +Aq`=;gr*d9σEsNv`~o? kهfy{6r֬gG| }`LDNOoҰ<XspKwFGxبnh?+.]%mh-=z_!lh2zgw& E`OЗNwC3o۫XeY>n+p#Pu&~K,vtA.u<~8?ݭNP y|\ ;d öOuuvpd#;IaAJ1sO>m[ $\vqdWi jgxKm&lzէ5m* ;tP\|p:(]֣qruYO([WaX9ce3ڰ[]J;<Ӑ3yXXF/=ZvDʱ}vt``[ 0Uk-YؑlЀ]ȳS)(N%aWAm~ =9BA<0Lh֓je㫥[׋ۖx^AWhB{TUg@/qɺZ[wba[9V Y2 Sqd,a"7P\XԙK\tl9CXn˗w"!WcXwK`2йx g;l^m=Cʂ8 Vx]p{ RYloAζʃ'(o] aϮ^l2ؼ;I`f]ǰ+7d ātR-Ypa9}i΂k7 ;`q]Xc?խ'^syrj¾ik@}ln+|!Pkf朰gҰ㲉'sbٺTtMR)ξ7ԃzlWcQ)vOփr:kM<_%KU"k ܬG.qۘEUoHz,SLgHª\!Z f`Պ<,؈-փBD=(cbʯ鬆Jj&GKvVnc+WgV]ƕa;@ڲA}Vc)`q&.Z J V׃i} mYj\qҧ) [kty"x{Qz z\lGbMld3ӖFש݃=>m Xa `MJilZmm} ldZ_֧kש}٬dӌdO[ ˸f3/ +S=>mX?Td˺קe;6%[Wm?=Cn9ϦtIENDB`}DyK _Ref530886833}DyK _Ref530829785{DyK  _Ref10887249 Dd x  C 4AafterorderingC"b i2 VVZr n i2 VVZrPNG  IHDRIsRGBPLTEg&T aIDATx] |`=q+|9ZKoNkW' <^ex|o\xP ׏#Z/ox9='羃z9YZ6J ;Zk?:1oZ}r  T) FyQ}441WTr7/]GQy?fD[n+- \?>z@v60ZWY`ǶTb/~_3aUʚEװDŁG \yjx7j/p̫ao.0u͋HҰWe $p]` IݒGF<<ݒT--w:~Ez1劅k,X/]}*yeKW[P >X8 =ӈ=V7[]%G5 2eԐ =?\d2$}%UV4na =c(hpȑVHl˜"E݀&yt0ݕ.fx2j5q!τ $ý |G]UGU5ҩS>gW]ŬJ3ƴ#kP?ՈxSjXY`E洤bӺxȜ˝Q_DU| - 51U׺,@xO38``VN bGnT?|=Ƨo#?2+8۽ʟMz}+hNyXU 4leȇPZAgv)x61vƹaF/Ecmn5l O# GM*rB )`Va7k7dG$9U 3`4 jïت~A6 ,g]5 V#5ֵ5 WT3 zg ݆+`F`x61'$5Z/Z=8~0.ZB} Mk0P~ȿ+2^q+њv1d٣1_F^--Pa$rƼr!&h}1S86VsWBNO5Ŀ*!R rxHhGPh6,@&^^mv|w,73zSؚQMٍ6/^Zځ8cz?[]t>7|VùS^ W71Ԃ_#6m-M Ae {/i15`xtĽ#2Ӧ/x!J&HxM  8 '7 2mpl_c*^q K=ۗ ˫PFޛIkp>1$Q=͒,+ڛIOKcjo+q^} `* =)Ia׭+4YѴw :P+pj+Û 1`p4b2匽t>{8I^miL4d/\VMl\e"xx:ڤPik-7g%~p1+, |7װonӅߦ&}+s;_kif1|03F?qsx#F]FYߗbu4*_WB/֖>Xe7j=IEu-l V\d#d;s7bG3bɖ`#j84'v=4> 6M:80̓MLF@1ȓ `.y8cV>Ca9@_i ?1UVOYoz7ݪM'<'2_?+υeHL GpG"{w; wcrtj&1v YLtbjωeu'T ъk|-5~4a楽67mՑ.3 \Lka@H$Qߚ.f9@B1Lg'2d ;E8s-LґNKF0q :]DfN;?fּ"7%]&&,ajMNgbHqKu1 _̍̿PbbL^>7jӤy;0ml/m2ٓwX_o[P؊\LLcLo69{VT.6{fp_ͼT]KHhص5C>؊~}pBu46~33ta[2}Dyf8-՛Tmxe[-0yyŝvN g{mn}lEoYuޔ_[S7weז)TfMM|B|81gL]|gBhiY^̜0+X8Eo>uSVz3R7e|WBZVq Ofnf~rN՛)v륔Z_Eo%fnʴGRaN9z=z3V7e5p̩R S{)5Sݔ.e=z61ܾBYyuSBRQMtz3қv_/(Kgw)_.3'Ec)Ojr'5gi̓T4gQL1NF&DՕmu1&vJ+c6cPLO gViDk(i'{ x&Sqs vOO;NI'ٯMM=xLp)3̌iݎPӟjeCW[)kT 9S[jfVc\ds65)nڟVjxV߁)^)nGu?cQ7$٘8+T8+y'+Ope1O1鵅C>ɋph'/9ٮώHfGb>uP3 `Jvx;f֫]*nӑs 3Q`Q밯zkq8~T[3Κ5;SePLk=@*n)Xjp.0 o=>&Z\ )1T38-y-ML}8n~x&vAA}|AASz!4-psKyyN>(t:?ዎ?; SJM<ƹKZ1)ztx`~labФYr!Mìa,u):mnJoH~]kqrn3)́co.t)Y3Isb'7DBFs+-ָm6'wK\`rKӘ8ms=S!_P"1Υ.8I֜s5_~$/b!} IENDB`}DyK _Ref530892005Dd p  C ,AafterpuncC" bY t}Yn5 n- t}YnPNG  IHDRsRGBPLTEg&T IDATx]0 a\asoeHҔRLQPLҷi|Bo_/eweq~tw);kws:߸ԵvmgTw0 gKg&s<80ìS% |T W4̺\ aNsm2Z|bc&5O1g4z)vs]m[j?}z40GULFT-rkAL cgSt镍)]L ˘keɦ34:!|D1[kcVd3rA/n 4 eGi𱵍~ t pʖF6CA/|wxô pLcwh~L|֘P,l9,6q^ ÃSbRsL\8 eI ? ֚ Є'v=-,wE0 l `J41y}˛mH~y} b.x aq^^[0iv=6pŢ7cS?ԧ8k쮦GX`:77 cm_oQ2 Sր]&뵉9i6NRa#VHΪƼkt)z̀Ռį)q9LIMjL 5֥lt vJo>"Umi!5b6~L,-XZ  U\3&2o33$@LDytuAF YG5b._kMy-|A!dקEs9EL٘sTRFk7N;XL\q1༈%v&}}*Fl.8AOgaE}K;1#JSS7f4j[Yf=-9F" V{^̘Tw^#jNjx@/Bwj !jI`MЏ9}ojdt2^#qfNBSKkLJcTƒu{$CzimT/qƂ0l l(]xٰ]?{7v< ^j.k2Cv譎jQ;Q s}[wA˭;s6S^WII"w=M G`8v7/1کmk}Ǐ|ƙymMü|LrRtiMĤ:Ri>zg|=?`!qnpj?s0qj@%Hٟ}7*kLZ xsL5N:R&j̫tAJW3UVFk6/(HܘX0qm F_P\3g |y/勹m_wrƙBs;0}3"N c.BLy O>5=:JwA̞0q@0}F{iLX&-(hޭ Ʃ1ibS^=}vcMu0.i>̝C0/6?Bg(|/L(Jޞvsql _y%܃iL N)u.q>O_1U S$N*f'/(m?^0ԸBNLo[wĭq{vSnmm)M$K 3/(q.Lu2OO!qF  :s{)fbޔqefh9*wJ_ JJ/M3G/fc3U.(R/\csfЧIx ;xjL8pT={&<*ەY-ϡ\xjY$8M͟p΍ >ߕ3Sp* OWܑ3Gԛ<*sf4 ff OSh3қ<)˾8v"y:cYp`'fb~l!ћiiy:}9sa8L[o7e3+SgL_PjNz7Xe䘠cpp8suIICfPC 3Xa6͇su¢?W L4^#?J@ZYUZSukI aV`ru&5y|(,__)VK/!WC|(4Ut,1EJ5cn=JYS忠anJ 8syZNnˍCS[mvsZm^Z9/[khD*Cԟy,&[md-sc8]^<%>D/yDCHKCVD@S1*JTӎȘ߲3u$k z&*-%i۶[abrی9I:P #bC;UyJLOcNN 3-[-MLUl:ҍ#- LԓY^wA9`!y,7' ?aj,N7sKi{/O.-]lbh}ȗҥAvj;&ZZ_Ì9|Yb~1_/Ŵ1gGgIENDB`}DyK _Ref530892005}DyK _Ref530901010 DdI y   C @A (recordforinflectionC" b |k#Zn  n |k#Zn PNG  IHDR'4sRGBPLTEf )IDATxQv* y ;hմ+׀ഝX6 iE!.茁A9wG8@.v>脂(<jȇ^(\0,`yKPN>l]x8T@ %XJRҩ J} 0L6҃h1:UfS5Krhէ_6'}6pQwpѮ A1(ŠPP,ae HSQU` ~,1ĻO I|R4djNY9S*87`CYI  Cik*-2PاoJg>$ 5Z/h]P:J^yvkj*](WTloP A1(#=ŁN(=fGPz ([=E1JO!ec=Ek)Bb7) 6POQf(]=HOQhz ]j] )bTlfP A1(ŠG"o6e@ÀŧTϊ@~0 WZ;,o)d!0TOH" (75P`P.hӼGpv*TbŠbP A1( ʟF.MaSeʨZ(i؎S| ; iRƝg>qRك̋J9x2> GP9>Et]XJ[9">edeJ3mhECK1(V A1(_BUI<@A6|3顬]H|DzROy>1c)Ƀ ۍGUrQGqȋ)k2Kyh! Ud&( t}3f.+ŠbP]prx*)Ăr.`4r G{\ >m(SҊOa M),ŠbP Ay(()w?s:Bg[9hV|SbVJ\F(6 n 7Mwk逡bJlJ_{(8N )q]ih[oŠbP Ayt(Amܯ{*K:ɹߖ=ϣ ۇP Sx}f'K|{&(}gy bʲ=s.n$VSw \+ nWWO_PƖ0ի6qZKJSf 嶁r,P|K(_o/ަ)?~9U)3R A1(ŊAM(n7Zq$/tCYʼX!-O i&s<< m9sr'mKɚ[mЍT@Lԅo嘥\ wڞ -N|Lɔ/a6.4{gbP A1(ѡ;Z3¡WUu{Pq>5c p[ߤ|6G|_ΧriNPYc9Dm9ſk-0>zButzsD)9 (c|q1SH'6qLB sDD7JMqW I (%駯^ Koki rŚ 7tE~^0qo|"L=Ѧ ,_U|^978_AYOs҈(wA)цg)ޯ-](oX F0XS]"smn (.CIH>{PcK]A\(ZotźʴKN+{JɩucSJ'Vb&pCߡ󑄓Os6F5;.?g ʣކܠbP A37#i9=q`Uy(h p iESQa)>Pݫ~}TTT@;څ|*߆BՄ> ϟ:[PTGO>yI I5 W|*6x3(ŠbP ʣC;iz|zZWl)a9|O (Mv}(R^? JHʺLXPL[RNEid=-ڧp <֖!%^o)m(׫"( EtbRĻ/Rb9v|\z r~(+ŠbP AyP(XZU*pv"~+]PbPţTk4D^̵GmA``Н^'E6ARa(Ń@5Jn$(肂|Kۇ"QbTY} AH1(ŠWPD&rmK.FŠbP A1(Š(.IENDB`}DyK _Ref530901197Dd^ p   C ,A inflectedC" bY N0D<'\|K5 n- N0D<'\|KPNG  IHDR6qsRGBPLTEg&T IDATx[ |` }Wk1\vtzoOyW`%^_-7<~B+jSUx78u-]۹wx7| ~@R#qg-i0~1U`ΗAjPUHì1>o6#iv-=G)̀6_VlwؤiaГۘ3v d;CKʁ.A~_=Z L:ZM *2acبaPnbzdL`]@]\+kLi:LMո.P&h E-:Pg@'&vB&Xx0{->z8&IOh~L|C00l1j`ygtt%0)WXPrӛ]ê\a 0=#6IT({Z)k0!~l82xv:{&VaU :X1a!tPoJFULTԙt}C47bYƖ# [0ոRi1 [H1qZf+%)悤>&Ё\ڥv+ؿ7קcֈ1֚DywoP07gLxy*E}+kO?bLS7s"=b]cQCQ kv jm>P?)(v8gӰ1s.^4Zq:/fz儂s~r?tM<,Q`V7`Bч_)n0},p&n}ӪmelZyҚʢU &y kD5Qw=-4iįnCJ:[9s1ODM]h zZ$GCKWZd6 Zc;-r5xt&[ORI%jU Zg[s0w'[ kLҖ0E>L*vezT a&4N0zZHZOkby1Shӆ&s0qj@'%H>KPNc5Zgꬅ5^. x'Zrc&j̻tAJWKZ-)vm(H16{aƩ2y>lgo^v/?qc~ (7tbo0Y_V҇afu1n}P j =j {fO8K>5&,j4aԘwW02nLt.PL܇st0} MW1=_P1? RmoOQ.)>M/Ua98O~A{0))6!2˹S3";7IŬQ<}ryƸ6u_MLOoİ=k)7ɺ6{_ޔ&K 3/(1q:Ob6r ōxS:/ENq0 í"m3RK-r=MҔ8O~Ai9cǻl\&/E&%f 3íOo'Iⶃ̄õ9R\3WS;9L5jʍT0ysJ2z's7Cy:w{~Qf(O0MbSot=^ϙ ڈaa60y:͞BK05Vf l՞BϙoqPtƲ 9aZ+Xoaa&̅iӛ}l# K9o3=?gVSgLǟP՞Nza=AՙOSWܘ|J]:sb,su|(A%`uAM>dJr \6vLȇb=`(^ժi:u ֔Xr.t&O'iw- $0#&K1ru*11Sύk)i j0Acsu PxblvD䜋ʺi:+Suko`$cMթk|(7bՒKP -ќVWwm)RpztuB94`nJp,Y10״ڼ0I?=,P΋Cԯy zd-[ Aty&y{3NVHM^GK L@S1AjZO;:Y#;֏`eXbrms)F#mhI]\L6c0IzOڻюR?.ݵ0KaB\>l70mw r%;VZ-MLVjo::rǟ)u~Mԕ^@Y?M߃ 0/i`TNsK)<Ǘy,KC 1<ƜvUV٣K<S(߆0-]ڄtѠ9E gk~=+љzQ3,={W󽋓u%}Ͼ;'nXi]=*L_ nM{OnK:S_lVoRGm6QO]Rudk?SwS_,5?ŧÕYMzjyXx.MU OJZGɚYo3|r8z;OOm,f2uf!f>xЛ˭Ժ*_roIqj19ORK+3੩|Qz3G u5<z[xuSonve?HQ칁듵fTz?~7y=iݞwxyk,Yrg?~&#6@mo{osr|L= _+=cos,<{ߦ~+Nۋx6s2'}t7s4vzl稼A=n[>yĦ'Nxhu[tH^s_~y&=_5w;Tؾ7v'{ywH85y49iAǶ4=e/u6!N~=ӎDB,÷}"c߳>icjKo}rq[wMl[/RWQSbN5gⶦʱ<+g.vx$<8w3W_\qQ<~m毈㹏pkyd5\>VVߩ1o? Uj1x^Ug?_^=1I֏DyOϷ_xc\z_~y?4:s3^ x:?IAN<[`}OuGk/9V{_ h;=c9cx\1||9e9573;fgg3˕U03;h5~(3mohuO__Һ zD6\$Յ:w_pʹms~w?ku1w8OzXW9w^>wlz'fZw!1ns罬3쿉?4o}?X9Ox~ g;Y{CO |UmXvzՍ\^ueً㵕rbW-_mxB"u߿~_$_J}Ks8և8S|v[5o9ӷ^9#^ͫ= \_+pʷTs*OmĜfK+\Ͼ|9ȥ]/?[֋Omk55SqԺY5eŒ,g5n-?su[`yU/U߭\ݹ &xr>&{OxS~>{!VVA1VRV;Y_k7xD]%9lͬ8~susO:&~o ;8lbMZ^Yy=~{-=U_Of+kܖ_t>dP"re͹l=ϱ_gH+]F|Gsk~w~׋NSY}HG,Q}^Q+^_:~ċMɞy?~@!;3P#_Jojzw=mvw~Ⱦ-_^YȂ_}5^%nFߕ^zυzG<.6uuCſ^ f&_MW͹vy^vG&=lj+1\^WNZj_=1}Awcwa9G%4^;*IƜɞ[9>NFgE| k1#^'|#JhN=}Lv{Bɗgm/h–pk5v*9_>N.Sk5U|Xh6j:/Wϧ=Ϸ4}T9es}rToQN-j>5_/|yoOh64ZڈL/nVuF7>o6lƍi0|%J·Ǩ+h7|CVo'fȠ=L>_Wm3 7>on3+4T͡qᛃɗw4F)W U|ϲ6Fh]25߮tS9 L M_#qшՈLo\o^Y _J+j__>؊.|[|٠/_^A_B k[hAk#[.b2 11)0|%J·t_?|E뇭 MПam Z;cK5/!f_BK }l Io9h!q4F)W U|=gYE#Vyua|p7>ovUX򗰆gm8Zh o\o^YD_"J+j__>؊.|["h |y/%D+k[hEk[".G|k"D5QfEo|%J Vvu_?|E뇭 MПam Z;u/X_,_,feIbi M74M_|% t+7}$_YfȢՃfޅ֛%>Sh-yʚ l,5{oy> ֦H67V:Gցw<{'8`u<Ʒ7P){,55V(^_3|5zj h͕Rih 6[?4[/Rg7ZR%g?!s͡C('Xo0;<yhELX`tOh333ૃm,:/Khգ/_ z_ZoYh@3w2R͗0"/RgEUdOhN?Q4F:LfWX灱ԙ@-eH||mK fjo{Y/ 34Th uͣ|1hH\Y~8z:x.ĬB|1 %Ko%l1!hm/ _$|`KI|BS/XC#!4\t/&01Y_lc8|ėk]htՎfڅd/A5_wY <2/A hu͓|IWX /2$_)RkC3w4𗀆|im0/ h uͣq5,/~fEo%| %[ [̅o/|K—/ RT 9ȡu~1 1ʐ `3\G'_ |%$_;]B v4.|'}iY~{_|'O_e(hF_|_-:c^ŋxː|mK ͤ Qo5_W'넯Ny t{%?4:Q4Օ?EcU ރWow=W8׿ 1|vzn&;'x_d W(B޾?Tůb+WݬCYgFowUrf*ZO+WmU]+9O?ssh"gAA]v+Q[P8q=َLzfWK0qۅz\]x9!g~Y\IW An y/A9\nWAtuޜFdg2#<y _ _wSN;  X mhm aaE8R[#9ﭑnj~#2_8_ ~ۼo8J+vߣJ^>ƃh=[/h-B:BDy!bƨQ-{E<Ȣ*(ZXBdr˃|}j^ A4Tm-h-|eȃV/Dc:#]m,ƍi0|%J·Ǩ+h7|CVo'fȠ=L>ݐZj~ffG~f*978|s43:eJ_^0YfȢuͼ n|oWYƨ=@tWY8khjDs]`&_77/,K /Af%5/W_lE>ݐA /_<|e5ֶтF4]' J1ʔEo|%J Vvc:E"|EV_ϰ6Fh1]%T/!%E[C4qof_|% t+7}$_YfȢՃfޅOd7%fGю0 K㬍Guaqk͋0_"KYI|E > +[хOd7%/×lD򗈚omEMhnOAT|k"Dh &ʬh W_ a.|{뇯_"|$>3͠Ak'{.b9YXXŬ,_,  &i7N2|eFė?,Yz̻<"CmMZː5O/_.|"!u|4|i wDch 9DD\Ynn?bViJ%KEPRINTCompObjLqObjInfoNVisioDocument      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~lsU^=f EMF8@F, EMF+@``FxEMF+0@?@ @ @;@^ @$/A9@23KC"A!b $$=='% % Ld P !??% % $$AAFEMF+@<0NnC@@H</A&B}/VC&B}/VC9@/A9@/A&B@$$==_8881% % V0#c c PP% % $$AA( FEMF+*@$33B%B/A9@@0$9=ARIAL6@0$II. Convert to Basic Tree`>""">9.>""">r>""">>""">>""">W?""">"?""">4?""">C?""">S?""">r^?""">UEh?""">r,p?""">Uy?""">?""">9v?""">?""">r?""">U?""">䀧?""">i?""">U]?""">U ?""">9?""">?""">??   RpArial%Monotype:Arial Bold:Versik%`h6`h`C`h`N`hB`X'<`wY8XG2 `$w Uwww w<P2 `M2`0dv% T0 AA0LII. Convert to Basic Tree% F@4EMF++@ @$oC&B4C&A$$==% % Ld :O !??% % $$AAFEMF+@<0NnC@@H<oCYlBGCYlBGC&BoC&BoCYlB@$$==_8881% % V0>II% % $$AA( FL@EMF+*@$33B%BoC&B@0$9=ARIAL6@basic_tree_harnessX>""">>""">9^?""">0?""">?""">'?""">7?""">3A?""">UEL?""">\?""">k?""">{?""">U?""">9v?""">?""">?""">?""">䀬?""">??   % T%z3AA1Lpbasic_tree_harness% F@4EMF++@ @$jCWC>C A$$==% % Ld!< !??% % $$AAFEMF+@<0NnC@@H<jCIBCXCIBCXCWCjCWCjCIBC@$$==_8881% % V0@% % $$AA( FthEMF+*@$33B%BjCWC@0$9=ARIAL6@introduce_coordinationr\>""">*>""">>""">r>""">>""">`?""">!?"""> 3?""">B?""">R?""">Ub?""">gr?""">?""">?""">r?""">r̘?""">?""">p?""">X?""">U?""">?""">举?""">??   % T '5AA 3Lxintroduce_coordination% F@4EMF++@ @$jCrC>C A$$==% % Ld ! !??% % $$AAFEMF+@<0NnC@@H<jCaCXCaCXCrCjCrCjCaC@$$==_8881% % V0!!% % $$AA( F$EMF+*@$33B%BjCrC@0$9=ARIAL6@update_parents?""">x?""">(?""">8:?"""> J?""">S?""">UUc?""">'s?""">C?""">r,?""">U?""">9?""">9N?"""> ?""">??   % T rAA Lhupdate_parents% FL@EMF++@ @ @$AC13KC0A$$==% % Ld !??% % $$AAFEMF+@ @H<NnC@@?@H<A&C\C&C\CCACA&C@$$==_888% %  ;m6GmX4Tm_x__TG6X4xmm=m6mX4mx6X4xmm=:m6kmX4ymxyk6:X4-"""x-m:m=m6mX4 mx 6X4xmm=_m6mX4mx6_X4RGGGxRm_m=m6"mX40m;x;;0"6X4xmm=m6mX4mx6X4vkkkxvmm=m6GmX4Tm_x__TG6X4xmm=m6mX4mx6X4xmm=;m6kmX4ymxyk6;X4-"""x-m;m=m6mX4 mx 6X4xmm=_m6mX4mx6_X4RGGGxRm_m=m6"mX40m;x;;0"6X4xmm=m6mX4mx6X4vkkkxvmm= m6G mX4T m_ x_ _ T G 6 X4 x m m= m6 mX4 m x    6 X4    x m m=; m6k mX4y m x  y k 6; X4- " " " x- m; m= m6 mX4 m x    6 X4    x m m=_ m6 mX4 m x    6_ X4R G G G xR m_ m= m6" mX40 m; x; ; 0 " 6 X4    x m m= m6 mX4 m x    6 X4v l l l xv m m= m6G mX4T m_ x_ _ T G 6 X4    x m m= mY( m  uX4 h ] ] ] h u6 X(   6 X4    x m m= 6 X4      6 X4 ! , , , ! = 6 PX4 C 8 8 8 C P6 X4      = Y(   X4      6 X(   6 X4      =R 6! X4     ! 6R X4_ j j j _ R = 6 X4 v v v   6 X4      =- 6 X4      6- X4; F F F ; - = 6j X4] R R R ] j 6 X4      = 6 X4      6 X4 ! ! !   =v 6F X48 - - - 8 F 6v X4     v = 6 X4      6 X4      =R 6! X4     ! 6R X4_ j j j _ R =6X4vvv6X4=-6X46-X4;EEE;-=6jX4]RRR]j6X4=6X46X4!!!=v6EX48---8E6vX4v=6X46X4=R6!X4!6RX4_jjj_R=6X4vvv6X4=-6X46-X4:EEE:-=6jX4\RRR\j6X4=6X46X4!!!=v6EX48---8E6vX4v=6X46X4=Q6!X4!6QX4_jjj_Q=.+6.[X4.i#ttti[6+X4#..+=.6.X4.#6X4#..=.O6.X4.#6OX4B77#7.B.O=<? % % $$AA( " FEMF+@ FEMF+@H<V@aCOCaCOCCV@CV@aC3@ *@$33B%V@aC@<0@> @$@>!b !% '% % Ld!??% % ( % " FEMF++@ @ 4@ @H<NnC@@?@H<V@aCOCaCOCCV@CV@aC@( $$==_888% '%  ;K6{X4{6KX4=222=K=6X4&&&6X4=o6X46oX4bWWWbo=62X4@KKK@26X4=6X46X4{{{=&6WX4dooodW6&X4&=6X46X4=K6|X4|6KX4=222=K=6X4&&&6X4=o6X46oX4bWWWbo=62X4@KKK@26X4=6X46X4|||=&6WX4eoooeW6&X4&=6X46X4=K6|X4|6KX4=333=K=6 X4 & & &   6X4=p 6 X4      6p X4b W W W b p = 63 X4@ K K K @ 3 6 X4      = 6 X4      6 X4 | | |   =& 6W X4e p p p e W 6& X4     & = 6 X4      6 X4      =K 6| X4     | 6K X4> 3 3 3 > K = Y(   X4      6 X(   6 X4      = ;6 X4      6 ;X4 H S S S H ;= 6 xX4 j _ _ _ j x6 X4      = Y(   X4      6 X(   6 X4 $ / / / $ = 6V X4H = = = H V 6 X4      = 6 X4      6 X4      =b 61 X4$    $ 1 6b X4o z z z o b = 6 X4      6 X4      == 6 X4      6= X4K V V V K = = 6z X4m b b b m z 6 X4      = 6X4 6 X4& 1 1 1 &  =6VX4H= ==HV6X4 =6X4 6X4    =b61X4$ $16bX4ozzz ob=6X4 6X4 ==6 X4  6=X4KVVV K==6zX4mb bbmz6X4 =6X4 6X4&111 &=6VX4H= ==HV6X4 =6X4 6X4    =b61X4$ $16bX4ozzz ob=6X4 6X4 ==6 X4  6=X4KVVV K==6zX4mb bbmz6X4 =6X4 6X4&111 &=6UX4H= ==HU6X4 =cR6cX4cXK=2262RX42D=9K9X9cDcR=c6cX4c"X-K-=-2"262X42=KXcc=cv6cX4cXK=2262vX42i=^K^X^cicv=<? % % $$AA( F`TEMF+*@$33B%BV@C@0$9=ARIAL6@III. Global Movementj>""">8>""">?""">j ?""">Q?""">8?""">X3?""">@;?""">L?""">^?""">m?""">u?""">}?""">䨊?""">X?""">A?""">*?""">Я?""">丷?""">h?""">??   % T3AA3LtIII. Global Movement   % F@4EMF++@ @$oCaC4C A( $$=='% % LdO !??% % $$AAFEMF+@H<NnC@@?@H<oCCGCCGCaCoCaCoCC@( $$==_888% '%  ;6X4'''6X4=p6X46pX4bWWWbp=63X4@KKK@36X4=0F6aFX4nFyQy^ylnvav60vX4#vl^Q#F0F=F6FX4F Q ^ lvv6vX4vl^QFF=UF6FX4FQ^lvv6UvX4Gv<l<^<QGFUF=F6FX4%F0Q0^0l%vv6vX4vl^QFF=yF6FX4FQ^lvv6yvX4lvala^aQlFyF= F6<FX4JFUQU^UlJv<v6 vX4vl^QF F=F6FX4FQ^lvv6vX4vl^QFF=0F6aFX4nFyQy^ylnvav60vX4#vl^Q#F0F=F6FX4F Q ^ lvv6vX4vl^QFF=UF6FX4FQ^lvv6UvX4Gv<l<^<QGFUF=F6FX4%F0Q0^0l%vv6vX4vl^QFF=yF6FX4FQ^lvv6yvX4lvala^aQlFyF= F6<FX4JFUQU^UlJv<v6 vX4vl^QF F=F6FX4FQ^lvv6vX4vl^QFF=0F6aFX4oFyQy^ylovav60vX4#vl^Q#F0F=F6FX4F Q ^ lvv6vX4vl^QFF=UF6FX4FQ^lvv6UvX4Gv=l=^=QGFUF=F6FX4%F0Q0^0l%vv6vX4vl^QFF=yF6FX4FQ^lvv6yvX4lvala^aQlFyF= F6=FX4JFUQU^UlJv=v6 vX4vl^QF F=1 61X41<IWbb6b X4bW"I"<"11 =1w61FX419<.I.W.b9bF6bwX4bWI<11w=161X41<IWbb6bX4bWI<11=6X46X4=]6,X4,6]X4kuuuk]=6X46X4=86X468X4FQQQF8=6uX4h]]]hu6X4=6X46X4!,,,!=6QX4C888CQ6X4=6X46X4=]6,X4,6]X4juuuj]=6X46X4=86X468X4FQQQF8=6uX4h]]]hu6X4=6X46X4!,,,!=6QX4C888CQ6X4=6X46X4=]6,X4,6]X4juuuj]=6X46X4=86X468X4FQQQF8=6uX4h]]]hu6X4=6X46X4!,,,!=<? % % $$AA( F|EMF+*@$33B%BoCaC@0$9=ARIAL6@$global_mvmt_harness_pull2>""">x>""">ы>""">Ǒ>""">Q>""">U>""">a?""">3?""">9~)?""">P9?""">R?""">\?""">Uk?""">UE}?""">?""">r?""">rĔ?""">U?""">9?""">?""">h?""">?""">?""">?""">??   % TAAL|global_mvmt_harness_pull  % F@4EMF++@ @$oCh D4C A( $$=='% % Ld(O !??% % $$AAFEMF+@H<NnC@@?@H<oC.'DGC.'DGCh DoCh DoC.'D@( $$==_888% '%  ;K(6|(X4(((((|(6K(X4>(3(3(3(>(K(=(6)X4)')')')))6(X4((((((=p)6)X4))))))6p)X4c)X)X)X)c)p)=0)6a)X4n)y)y)y)n)a)60)X4#))))#)0)=)6)X4) ) ) )))6)X4))))))=U)6)X4))))))6U)X4G)<)<)<)G)U)=)6)X4%)0)0)0)%))6)X4))))))=y)6)X4))))))6y)X4l)a)a)a)l)y)= )6<)X4J)U)U)U)J)<)6 )X4))))) )=)6)X4))))))6)X4))))))=0)6a)X4n)y)y)y)n)a)60)X4#))))#)0)=)6)X4) ) ) )))6)X4))))))=U)6)X4))))))6U)X4G)<)<)<)G)U)=)6)X4%)0)0)0)%))6)X4))))))=y)6)X4))))))6y)X4l)a)a)a)l)y)= )6<)X4J)U)U)U)J)<)6 )X4))))) )=)6)X4))))))6)X4))))))=0)6a)X4o)y)y)y)o)a)60)X4#))))#)0)=)6)X4) ) ) )))6)X4))))))=U)6)X4))))))6U)X4G)=)=)=)G)U)=)6)X4%)0)0)0)%))6)X4))))))=y)6)X4))))))6y)X4l)a)a)a)l)y)= )6=)X4J)U)U)U)J)=)6 )X4))))) )=1w)61G)X419)<.)I.)W.)b9)bG)6bw)X4b)W)I)<)1)1w)=1(61(X41(<(I(W(b(b(6b(X4b(W(I(<(1(1(=1S(61"(X41(< (I (W (b(b"(6bS(X4b`(Wk(Ik(<k(1`(1S(=3(63(X43(((( (((6(X4( ((((3(3(=]3(6,3(X43(((( ((,(6](X4k(u (u(u((k3(]3(=3(63(X43(((( (((6(X4( ((((3(3(=83(63(X43(((( (((68(X4F(Q (Q(Q((F3(83(=3(6u3(X4h3(]((](] (h(u(6(X4( ((((3(3(=3(63(X43(((( (((6(X4!(, (,(,((!3(3(=3(6Q3(X4C3(8((8(8 (C(Q(6(X4( ((((3(3(=3(63(X43(((( (((6(X4( ((((3(3(=]3(6,3(X43(((( ((,(6](X4j(u (u(u((j3(]3(=3(63(X43(((( (((6(X4( ((((3(3(=83(63(X43(((( (((68(X4F(Q (Q(Q((F3(83(=3(6u3(X4h3(]((](] (h(u(6(X4( ((((3(3(=3(63(X43(((( (((6(X4!(, (,(,((!3(3(=3(6Q3(X4C3(8((8(8 (C(Q(6(X4( ((((3(3(=3(63(X43(((( (((6(X4( ((((3(3(=]3(6,3(X43(((( ((,(6](X4j(u (u(u((j3(]3(=3(63(X43(((( (((6(X4( ((((3(3(=83(63(X43(((( (((68(X4F(Q (Q(Q((F3(83(=3(6u3(X4h3(]((](] (h(u(6(X4( ((((3(3(=3(63(X43(((( (((6(X4!(, (,(,((!3(3(=<?~ % % $$AA( F|EMF+*@$33B%BoCh D@0$9=ARIAL6@$global_mvmt_harness_push>""">@U>""">rt>""">9.>""">9>""">Ǒ>""">`>""">ǁ?""">r ?""">90?""">I?""">aS?""">3c?""">t?""">2?""">?""">k?""">rT?""">U=?""">9&?""">?""">?""">o?""">X?""">??   % TAAL|global_mvmt_harness_push  % F@4EMF++@ @$oC;D4C A( $$=='% % Ld.O !??% % $$AAFEMF+@H<NnC@@?@H<oCFBDGCFBDGC;DoC;DoCFBD@( $$==_888% '%  ;/6B/X4P/[/[/[/P/B/6/X4/...//=/6/X4//////6/X4//////=606g0X4t0000t0g0660X4)0000)060=0z06az0X4nz0y0y0y0n0a0600X4#0000#z00z0=z06z0X4z0 0 0 00060X40000z0z0=Uz06z0X4z0000006U0X4G0<0<0<0Gz0Uz0=z06z0X4%z0000000%0060X40000z0z0=yz06z0X4z0000006y0X4l0a0a0a0lz0yz0= z06<z0X4Jz0U0U0U0J0<06 0X40000z0 z0=z06z0X4z00000060X40000z0z0=0z06az0X4nz0y0y0y0n0a0600X4#0000#z00z0=z06z0X4z0 0 0 00060X40000z0z0=Uz06z0X4z0000006U0X4G0<0<0<0Gz0Uz0=z06z0X4%z0000000%0060X40000z0z0=yz06z0X4z0000006y0X4l0a0a0a0lz0yz0= z06<z0X4Jz0U0U0U0J0<06 0X40000z0 z0=z06z0X4z00000060X40000z0z0=0z06az0X4oz0y0y0y0o0a0600X4#0000#z00z0=z06z0X4z0 0 0 00060X40000z0z0=Uz06z0X4z0000006U0X4G0=0=0=0Gz0Uz0=z06z0X4%z0000000%0060X40000z0z0=yz06z0X4z0000006y0X4l0a0a0a0lz0yz0= z06=z0X4Jz0U0U0U0J0=06 0X40000z0 z0=1=061 0X41/</I/W/b/b 06b=0X4bK0WV0IV0<V01K01=0=1/61z/X41m/<b/Ib/Wb/bm/bz/6b/X4b/W/I/</1/1/=1/61.X41.<.I.W.b.b.6b/X4b&/W1/I1/<1/1&/1/=.6.X4......6.X4......=].6,.X4.....,.6].X4k.u.u.u.k.].=.6.X4......6.X4......=8.6.X4......68.X4F.Q.Q.Q.F.8.=.6u.X4h.].].].h.u.6.X4......=.6.X4......6.X4!.,.,.,.!..=.6Q.X4C.8.8.8.C.Q.6.X4......=.6.X4......6.X4......=].6,.X4.....,.6].X4j.u.u.u.j.].=.6.X4......6.X4......=8.6.X4......68.X4F.Q.Q.Q.F.8.=.6u.X4h.].].].h.u.6.X4......=.6.X4......6.X4!.,.,.,.!..=.6Q.X4C.8.8.8.C.Q.6.X4......=.6.X4......6.X4......=].6,.X4.....,.6].X4j.u.u.u.j.].=.6.X4......6.X4......=8.6.X4......68.X4F.Q.Q.Q.F.8.=.6u.X4h.].].].h.u.6.X4......=.6.X4......6.X4!.,.,.,.!..=<?  % % $$AA( FL@EMF+*@$33B%BoC;D@0$9=ARIAL6@set_agreement_bitsѼ>""">Uu>""">>""">U?""">W?""">('?""">8?""">C?""">rlS?""">9>c?""">|?""">U-?""">Uݎ?""">Ǚ?""">?""">2?""">9&?""">?""">??   % T|AALpset_agreement_bits % F@4EMF++@ @$oC-D4C A( $$=='% % Ld~+O !??% % $$AAFEMF+@H<NnC@@?@H<oC4DGC4DGC-DoC-DoC4D@( $$==_888% '%  ;+6+X4++++++6+X4++++++=A,6r,X4,,,,,r,6A,X43,(,(,(,3,A,=,6-X4------6,X4,,,,,,=0-6a-X4n-y"-y/-y=-nH-aH-60H-X4#H-=-/-"-#-0-=-6-X4- "- /- =-H-H-6H-X4H-=-/-"---=U-6-X4-"-/-=-H-H-6UH-X4GH-<=-</-<"-G-U-=-6-X4%-0"-0/-0=-%H-H-6H-X4H-=-/-"---=y-6-X4-"-/-=-H-H-6yH-X4lH-a=-a/-a"-l-y-= -6<-X4J-U"-U/-U=-JH-<H-6 H-X4H-=-/-"-- -=-6-X4-"-/-=-H-H-6H-X4H-=-/-"---=0-6a-X4n-y"-y/-y=-nH-aH-60H-X4#H-=-/-"-#-0-=-6-X4- "- /- =-H-H-6H-X4H-=-/-"---=U-6-X4-"-/-=-H-H-6UH-X4GH-<=-</-<"-G-U-=-6-X4%-0"-0/-0=-%H-H-6H-X4H-=-/-"---=y-6-X4-"-/-=-H-H-6yH-X4lH-a=-a/-a"-l-y-= -6<-X4J-U"-U/-U=-JH-<H-6 H-X4H-=-/-"-- -=-6-X4-"-/-=-H-H-6H-X4H-=-/-"---=0-6a-X4o-y"-y/-y=-oH-aH-60H-X4#H-=-/-"-#-0-=-6-X4- "- /- =-H-H-6H-X4H-=-/-"---=U-6-X4-"-/-=-H-H-6UH-X4GH-==-=/-="-G-U-=-6-X4%-0"-0/-0=-%H-H-6H-X4H-=-/-"---=y-6-X4-"-/-=-H-H-6yH-X4lH-a=-a/-a"-l-y-= -6=-X4J-U"-U/-U=-JH-=H-6 H-X4H-=-/-"-- -=1,61,X41,<,I,W,b,b,6b,X4b,W,I,<,1,1,=1H,61,X41 ,<+I+W+b ,b,6bH,X4bV,Wa,Ia,<a,1V,1H,=1+61+X41x+<m+Im+Wm+bx+b+6b+X4b+W+I+<+1+1+=+6+X4++~+p+e+e+6e+X4e+p+~++++=]+6,+X4++~+p+e+,e+6]e+X4ke+up+u~+u+k+]+=+6+X4++~+p+e+e+6e+X4e+p+~++++=8+6+X4++~+p+e+e+68e+X4Fe+Qp+Q~+Q+F+8+=+6u+X4h+]+]~+]p+he+ue+6e+X4e+p+~++++=+6+X4++~+p+e+e+6e+X4!e+,p+,~+,+!++=+6Q+X4C+8+8~+8p+Ce+Qe+6e+X4e+p+~++++=+6+X4++~+p+e+e+6e+X4e+p+~++++=]+6,+X4++~+p+e+,e+6]e+X4je+up+u~+u+j+]+=+6+X4++~+p+e+e+6e+X4e+p+~++++=8+6+X4++~+p+e+e+68e+X4Fe+Qp+Q~+Q+F+8+=+6u+X4h+]+]~+]p+he+ue+6e+X4e+p+~++++=+6+X4++~+p+e+e+6e+X4!e+,p+,~+,+!++=+6Q+X4C+8+8~+8p+Ce+Qe+6e+X4e+p+~++++=+6+X4++~+p+e+e+6e+X4e+p+~++++=]+6,+X4++~+p+e+,e+6]e+X4je+up+u~+u+j+]+=+6+X4++~+p+e+e+6e+X4e+p+~++++=8+6+X4++~+p+e+e+68e+X4Fe+Qp+Q~+Q+F+8+=+6u+X4h+]+]~+]p+he+ue+6e+X4e+p+~++++=+6+X4++~+p+e+e+6e+X4!e+,p+,~+,+!++=<? % % $$AA( FEMF+*@$33B%BoC-D@0$9=ARIAL6@0$distribute_vform_features*R>""">UՋ>""">>""">G>""">8>""">r\>""">*>""">Uu ?""">U?""">9N&?"""> 6?""">E?""">U?""">r<_?""">rp?""">9{?""">r|?""">Ue?""">!?"""> ?""">?""">?""">`?""">?""">?""">??   % TAALdistribute_vform_features % F@4EMF++@ @$dCYlB4C&A( $$=='% % Ld;oUP !??% % $$AAFEMF+@H<NnC@@?@H<dCrBDrBDYlBdCYlBdCrB@( $$==_888% '%  ;6X4",,,"6X4=v6X46vX4h]]]hv=69X4FQQQF96X4=L6LX4!L,W,d,r!||6|X4|rdWLL=vL6LX4LWdr||6v|X4h|]r]d]WhLvL=L69LX4FLQWQdQrF|9|6|X4|rdWLL=L6LX4LWdr||6|X4|rdWLL=,L6]LX4kLvWvdvrk|]|6,|X4|rdWL,L=L6LX4LWdr||6|X4|rdWLL=QL6LX4LWdr||6Q|X4D|9r9d9WDLQL=L6 LX4" L- W- d- r" | |6|X4|rdWLL=v L6 LX4 L W d r | |6v |X4h |] r] d] Wh Lv L=!L69!LX4F!LQ!WQ!dQ!rF!|9!|6!|X4 | r d W L!L=!L6!LX4!L!W!d!r!|!|6!|X4!|!r!d!W!L!L=-"L6]"LX4k"Lv"Wv"dv"rk"|]"|6-"|X4"|"r"d"W"L-"L="L6"LX4"L#W#d#r"|"|6"|X4"|"r"d"W"L"L=Q#L6#LX4#L#W#d#r#|#|6Q#|X4D#|9#r9#d9#WD#LQ#L=#L6$LX4"$L-$W-$d-$r"$|$|6#|X4#|#r#d#W#L#L=v$L6$LX4$L$W$d$r$|$|6v$|X4h$|]$r]$d]$Wh$Lv$L=%L69%LX4F%LQ%WQ%dQ%rF%|9%|6%|X4$|$r$d$W$L%L=%L6%LX4%L%W%d%r%|%|6%|X4%|%r%d%W%L%L=-&L6]&LX4k&Lv&Wv&dv&rk&|]&|6-&|X4&|&r&d&W&L-&L=&L6&LX4&L'W'd'r&|&|6&|X4&|&r&d&W&L&L=&6&X4&&& '''6'X4' '(&(&(&&=&}6&LX4&?&4&4 '4'?'L6'}X4' '&&&&}=&6&X4&&& '''6'X4' '&&&&=&6r&X4d&Y&Y&Y&d&r&6&X4&&&&&&=&6%X4%%%%%%6&X4&)&)&)&&&=~%6M%X4@%5%5%5%@%M%6~%X4%%%%%~%=$6$X4$$$$$$6$X4$%%%$$=Y$6)$X4$$$$$)$6Y$X4g$r$r$r$g$Y$=#6#X4#~#~#~###6#X4######=5#6#X4"""""#65#X4B#M#M#M#B#5#="6r"X4d"Y"Y"Y"d"r"6"X4""""""="6!X4!!!!!!6"X4"("("("""=~!6M!X4@!5!5!5!@!M!6~!X4!!!!!~!= 6 X4      6 X4 !!!  =Y 6( X4     ( 6Y X4g r r r g Y =6X4~~~6X4=56X465X4BMMMB5=6rX4dYYYdr6X4=6X46X4(((=~6MX4?555?M6~X4~=6X46X4=Y6(X4(6YX4gqqqgY=6X46X4=<?8sZ % % $$AA( F0$EMF+*@$33B%BdCYlB@0$9=ARIAL6@remove_subj_obj >""">9.>"""> ?""">3#?""">4?""">UeD?""">7T?""">d?""">s?""">U?""">UM?""">@?""">)?""">ٟ?""">lj?""">??   % T@@NAALLlremove_subj_obj % F@4EMF++@ @$dCB4C$A( $$=='% % LdkoP !??% % $$AAFEMF+@<0NnC@@H<dCKCDKCDBdCBdCKC@$$==_8881% % V0hs[&[&[% % $$AA( FEMF+*@$33B%BdCB@0$9=ARIAL6@xlf2basic/?""">6?""">h@?""">:P?""">a?""">rlq?""">?""">?""">??   % T|p(~AA|L\lf2basic% F@4EMF++@ @$dC7C4C A$$==% % Ldo P !??% % $$AAFEMF+@<0NnC@@H<dC 5CD 5CD7CdC7CdC 5C@$$==_8881% % V0sQ &Q &  Q % % $$AA( FEMF+*@$33B%BdC7C@0$9=ARIAL6@ adjust_labels ?""">U?""">U+?""">r2?""">r\D?""">9.T?""">]?""">xm?""">`u?""">䘂?""">H?""">1?""">U%?""">??   % T7AA Lhadjust_labels% F@4EMF++@ @$dC_IC4C A$$==% % Ldo P !??% % $$AAFEMF+@<0Ne>@H<dCvdCDvdCD_ICdC_ICdCvdC@$$==_888% % V0qH&H&  H% % $$AA( FEMF+*@$33B%BdC_IC@0$9=ARIAL6@ split_sepfixU?""">9+?""">;?""">B?""">SH?""">:P?""">r `?""">UEn?""">~?""">r?""">?""">?""">??   ( RpArialMonotype:Arial Regular:Vek%`h6`h`C`h hB`wTwwpww @ w `$wwwww@ 53@ -w w< !dv% T0AA Ldsplit_sepfix% F@4EMF++@ @$dCC4C A$$==% % Ldo^P !??% % $$AAFEMF+@H<NnC@@?@H<dCs<DDs<DDCdCCdCs<D@( $$==_888% '%  ;6X46X4vvv=!6RX4_jjj_R6!X4   !=6X46X4=6X4!, , , !( ( 6( X4(    =v6X4   ( ( 6v( X4h( ] ] ] hv=69X4FQ Q Q F( 9( 6( X4(    =6X4   ( ( 6( X4(    =,6]X4kv v v k( ]( 6,( X4(    ,=6X4   ( ( 6( X4(    =Q6X4   ( ( 6Q( X4D( 9 9 9 DQ=6 X4" -  -  -  " (  ( 6( X4(    =v 6 X4     ( ( 6v ( X4h ( ]  ]  ]  h v =!69!X4F!Q! Q! Q! F!( 9!( 6!( X4 (    !=!6!X4!! ! ! !( !( 6!( X4!( ! ! ! !!=-"6]"X4k"v" v" v" k"( ]"( 6-"( X4"( " " " "-"="6"X4"# # # "( "( 6"( X4"( " " " ""=Q#6#X4## # # #( #( 6Q#( X4D#( 9# 9# 9# D#Q#=#6$X4"$-$ -$ -$ "$( $( 6#( X4#( # # # ##=v$6$X4$$ $ $ $( $( 6v$( X4h$( ]$ ]$ ]$ h$v$=%69%X4F%Q% Q% Q% F%( 9%( 6%( X4$( $ $ $ $%=%6%X4%% % % %( %( 6%( X4%( % % % %%=-&6]&X4k&v& v& v& k&( ]&( 6-&( X4&( & & & &-&=&6&X4&' ' ' &( &( 6&( X4&( & & & &&=&6&X4&}&r&r 'r'}'6'X4' '&&&&=&)6&X4&&& '''6')X4'6 'A&A&A&6&)=&6&fX4&X&M&M 'M'X'f6'X4' '&&&&=&v6r&vX4d&vY&lY&^Y&Qd&Fr&F6&FX4&F&Q&^&l&v&v=&v6%vX4%v%l%^%Q%F%F6&FX4&F)&Q)&^)&l&v&v=~%v6M%vX4@%v5%l5%^5%Q@%FM%F6~%FX4%F%Q%^%l%v~%v=$v6$vX4$v$l$^$Q$F$F6$FX4$F%Q%^%l$v$v=Y$v6)$vX4$v$l$^$Q$F)$F6Y$FX4g$Fr$Qr$^r$lg$vY$v=#v6#vX4#v~#l~#^~#Q#F#F6#FX4#F#Q#^#l#v#v=5#v6#vX4"v"l"^"Q"F#F65#FX4B#FM#QM#^M#lB#v5#v="v6r"vX4d"vY"lY"^Y"Qd"Fr"F6"FX4"F"Q"^"l"v"v="v6!vX4!v!l!^!Q!F!F6"FX4"F("Q("^("l"v"v=~!v6M!vX4@!v5!l5!^5!Q@!FM!F6~!FX4!F!Q!^!l!v~!v= v6 vX4 v l ^ Q F F6 FX4 F!Q!^!l v v=Y v6( vX4 v l ^ Q F( F6Y FX4g Fr Qr ^r lg vY v=v6vX4v~l~^~QFF6FX4FQ^lvv=5v6vX4vl^QFF65FX4BFMQM^MlBv5v=v6rvX4dvYlY^YQdFrF6FX4FQ^lvv=v6vX4vl^QFF6FX4F(Q(^(lvv=~v6MvX4?v5l5^5Q?FMF6~FX4FQ^lv~v=v6vX4vl^QFF6FX4FQ^lvv=Yv6(vX4vl^QF(F6YFX4gFqQq^qlgvYv=v6vX4vl^QFF6FX4FQ^lvv=<?s % % $$AA( F8,EMF+*@$33B%BdCC@0$9=ARIAL6@raising_movement>""">>""">7?""">9 ?""">?"""> ?""">72?""">C?""">hS?""">l?""">~?""">?""">ێ?""">䀛?""">i?""">?""">??   ( RpArial%Monotype:Arial Bold:Versik%`h6`h`C`h hB`wTwwpww @ w `$wwwww@ 53@ -w w< !dv% TCAALlraising_movement  % F@4EMF++@ @$dClD4C A( $$=='% % Ldo6!P !??% % $$AAFEMF+@H<NnC@@?@H<dCe DDe DDlDdClDdCe D@( $$==_888% '%  ;!6#"X40";";";"0"#"6!X4!!!!!!="6"X4""""""6"X4w"l"l"l"w""=#6G#X4U#`#`#`#U#G#6#X4 #""" ##=Z#6Z#X4!Z#,e#,s#,#!##6#X4##s#e#Z#Z#=vZ#6Z#X4Z#e#s####6v#X4h#]#]s#]e#hZ#vZ#=Z#69Z#X4FZ#Qe#Qs#Q#F#9#6#X4##s#e#Z#Z#=Z#6Z#X4Z#e#s####6#X4##s#e#Z#Z#=,Z#6]Z#X4kZ#ve#vs#v#k#]#6,#X4##s#e#Z#,Z#=Z#6Z#X4Z#e#s####6#X4##s#e#Z#Z#=QZ#6Z#X4Z#e#s####6Q#X4D#9#9s#9e#DZ#QZ#=Z#6 Z#X4" Z#- e#- s#- #" # #6#X4##s#e#Z#Z#=v Z#6 Z#X4 Z# e# s# # # #6v #X4h #] #] s#] e#h Z#v Z#=!Z#69!Z#X4F!Z#Q!e#Q!s#Q!#F!#9!#6!#X4 # # s# e# Z#!Z#=!Z#6!Z#X4!Z#!e#!s#!#!#!#6!#X4!#!#!s#!e#!Z#!Z#=-"Z#6]"Z#X4k"Z#v"e#v"s#v"#k"#]"#6-"#X4"#"#"s#"e#"Z#-"Z#="Z#6"Z#X4"Z##e##s###"#"#6"#X4"#"#"s#"e#"Z#"Z#=Q#Z#6#Z#X4#Z##e##s#######6Q##X4D##9##9#s#9#e#D#Z#Q#Z#=#Z#6$Z#X4"$Z#-$e#-$s#-$#"$#$#6##X4#####s##e##Z##Z#=v$Z#6$Z#X4$Z#$e#$s#$#$#$#6v$#X4h$#]$#]$s#]$e#h$Z#v$Z#=%Z#69%Z#X4F%Z#Q%e#Q%s#Q%#F%#9%#6%#X4$#$#$s#$e#$Z#%Z#=%Z#6%Z#X4%Z#%e#%s#%#%#%#6%#X4%#%#%s#%e#%Z#%Z#=-&Z#6]&Z#X4k&Z#v&e#v&s#v&#k&#]-&#X4&#&#&s#&e#&Z#-&Z#=&Z#6&Z#X4&Z#'e#'s#'#&#&#X4&#&#&s#&e#&Z#&Z#=&#6&"X4&"&"&" '"'"'"6'#X4'+# '6#&6#&6#&+#&#=&"6&["X4&M"&B"&B" 'B"'M"'["6'"X4'" '"&"&"&"&"=&!6&!X4&!&!&! '!'!'!6'!X4'" '"&"&"&"&!=&!6r&!X4d&!Y&!Y&!Y&!d&!r&!6&!X4&!&!&!&!&!&!=&!6%!X4%!%!%!%!%!%!6&!X4&!)&!)&!)&!&!&!=~%!6M%!X4@%!5%!5%!5%!@%!M%!6~%!X4%!%!%!%!%!~%!=$!6$!X4$!$!$!$!$!$!6$!X4$!%!%!%!$!$!=Y$!6)$!X4$!$!$!$!$!)$!6Y$!X4g$!r$!r$!r$!g$!Y$!=#!6#!X4#!~#!~#!~#!#!#!6#!X4#!#!#!#!#!#!=5#!6#!X4"!"!"!"!"!#!65#!X4B#!M#!M#!M#!B#!5#!="!6r"!X4d"!Y"!Y"!Y"!d"!r"!6"!X4"!"!"!"!"!"!="!6!!X4!!!!!!!!!!!!6"!X4"!("!("!("!"!"!=~!!6M!!X4@!!5!!5!!5!!@!!M!!6~!!X4!!!!!!!!!!~!!= !6 !X4 ! ! ! ! ! !6 !X4 !!!!!!! ! !=Y !6( !X4 ! ! ! ! !( !6Y !X4g !r !r !r !g !Y !=!6!X4!~!~!~!!!6!X4!!!!!!=5!6!X4!!!!!!65!X4B!M!M!M!B!5!=!6r!X4d!Y!Y!Y!d!r!6!X4!!!!!!=!6!X4!!!!!!6!X4!(!(!(!!!=~!6M!X4?!5!5!5!?!M!6~!X4!!!!!~!=!6!X4!!!!!!6!X4!!!!!!=Y!6(!X4!!!!!(!6Y!X4g!q!q!q!g!Y!=!6!X4!!!!!!6!X4!!!!!!=<?s; % % $$AA( FEMF+*@$33B%BdClD@0$9=ARIAL6@ wh_movementU5?""">UU?""">U(?""">8?""">Q?""">1c?""">s?""">j?""">?""">?""">䨞?""">??   % T!:/AA- Ldwh_movement   % F@4EMF++@ @$dC_D4C A( $$=='% % LdRol$%P !??% % $$AAFEMF+@H<NnC@@?@H<dCXTDDXTDD_DdC_DdCXTD@( $$==_888% '%  ;U%6%X4%%%%%%6U%X4G%<%<%<%G%U%=%6&X4%&0&0&0&%&&6%X4%%%%%%=y&6&X4&&&&&&6y&X4l&a&a&a&l&y&=&6&X4!&,&,&,&!&&6&X4&&&&&&=v&6&X4&&&&&&6v&X4h&]&]&]&h&v&=&69&X4F&Q&Q&Q&F&9&6&X4&&&&&&=&6&X4&&&&&&6&X4&&&&&&=,&6]&X4k&v&v&v&k&]&6,&X4&&&&&,&=&6&X4&&&&&&6&X4&&&&&&=Q&6&X4&&&&&&6Q&X4D&9&9&9&D&Q&=&6 &X4" &- &- &- &" & &6&X4&&&&&&=v &6 &X4 & & & & & &6v &X4h &] &] &] &h &v &=!&69!&X4F!&Q!&Q!&Q!&F!&9!&6!&X4 & & & & &!&=!&6!&X4!&!&!&!&!&!&6!&X4!&!&!&!&!&!&=-"&6]"&X4k"&v"&v"&v"&k"&]"&6-"&X4"&"&"&"&"&-"&="&6"&X4"&#&#&#&"&"&6"&X4"&"&"&"&"&"&=Q#&6#&X4#&#&#&#&#&#&6Q#&X4D#&9#&9#&9#&D#&Q#&=#&6$&X4"$&-$&-$&-$&"$&$&6#&X4#&#&#&#&#&#&=v$&6$&X4$&$&$&$&$&$&6v$&X4h$&]$&]$&]$&h$&v$&=%&69%&X4F%&Q%&Q%&Q%&F%&9%&6%&X4$&$&$&$&$&%&=%&6%&X4%&%&%&%&%&%&6%&X4%&%&%&%&%&%&=-&&6]&&X4k&&v&&v&&v&&k&&]&&6-&&X4&&&&&&&&&&-&&=&&6&&X4&&'&'&'&&&&&6&&X4&&&&&&&&&&&&=&&6&P&X4&C&&8&&8& '8&'C&'P&6'&X4'& '&&&&&&&&&=&%6&%X4&%&%&% '%'%'%6'%X4'% '&&&&&&%&%=&\%6&,%X4&%&%&% '%'%',%6'\%X4'j% 'u%&u%&u%&j%&\%=&<%6r&<%X4d&<%Y&2%Y&$%Y&%d& %r& %6& %X4& %&%&$%&2%&<%&<%=&<%6%<%X4%<%%2%%$%%%% %% %6& %X4& %)&%)&$%)&2%&<%&<%=~%<%6M%<%X4@%<%5%2%5%$%5%%@% %M% %6~% %X4% %%%%$%%2%%<%~%<%=$<%6$<%X4$<%$2%$$%$%$ %$ %6$ %X4$ %%%%$%%2%$<%$<%=Y$<%6)$<%X4$<%$2%$$%$%$ %)$ %6Y$ %X4g$ %r$%r$$%r$2%g$<%Y$<%=#<%6#<%X4#<%~#2%~#$%~#%# %# %6# %X4# %#%#$%#2%#<%#<%=5#<%6#<%X4"<%"2%"$%"%" %# %65# %X4B# %M#%M#$%M#2%B#<%5#<%="<%6r"<%X4d"<%Y"2%Y"$%Y"%d" %r" %6" %X4" %"%"$%"2%"<%"<%="<%6!<%X4!<%!2%!$%!%! %! %6" %X4" %("%("$%("2%"<%"<%=~!<%6M!<%X4@!<%5!2%5!$%5!%@! %M! %6~! %X4! %!%!$%!2%!<%~!<%= <%6 <%X4 <% 2% $% % % %6 %X4 %!%!$%!2% <% <%=Y <%6( <%X4 <% 2% $% % %( %6Y %X4g %r %r $%r 2%g <%Y <%=<%6<%X4<%~2%~$%~% % %6 %X4 %%$%2%<%<%=5<%6<%X4<%2%$%% % %65 %X4B %M%M$%M2%B<%5<%=<%6r<%X4d<%Y2%Y$%Y%d %r %6 %X4 %%$%2%<%<%=<%6<%X4<%2%$%% % %6 %X4 %(%($%(2%<%<%=~<%6M<%X4?<%52%5$%5%? %M %6~ %X4 %%$%2%<%~<%=<%6<%X4<%2%$%% % %6 %X4 %%$%2%<%<%=Y<%6(<%X4<%2%$%% %( %6Y %X4g %q%q$%q2%g<%Y<%=<%6<%X4<%2%$%% % %6 %X4 %%$%2%<%<%=<?Osq % % $$AA( F`TEMF+*@$33B%BdC_D@0$9=ARIAL6@relativizer_movement>""">r<>""">>""">9>""">(?""">ǡ?""">?""">Z&?""">A.?""">zrLL?""">9^W?""">0g?""">U=?""">U?""">9֐?""">?""">rd?""">UM?""">U?""">??   % TWKeAAcLtrelativizer_movement  % " F(EMF++@ @ FEMF+@H<@ODSCODSC ID@ ID@OD3@ *@$33B%@OD@<0@> @$@>!b !%@% '% % Ld%?$!??% % ( % " FEMF++@ 4@ @<0NnC@@H<@ODSCODSC ID@ ID@OD@$$==_8881% % W0!B}31 31 D2}D2}3% % $$AA( FEMF+*@$33B%B@ ID@0$9=ARIAL6@`TIV. Intra Constituent Orderingc>""">Ǒ>""">>""">Q>"""> >""">9>""">9>""">P?""">a?""">3!?""">)?""">U=?""">UO?""">Ue`?""">7p?""">y?""">ˀ?""">?""">8?"""> ?""">О?""">U?""">䀧?""">䐲?""">?""">?""">?""">;?""">/?""">?""">??   % T)7AA5LIV. Intra Constituent Ordering % F@4EMF++@ @$oCOD4C A( $$=='% % Ld?Y3O !??% % $$AAFEMF+@<0NnC@@H<oCVDGCVDGCODoCODoCVD@$$==_8881% % V0<]5I5I335% % $$AA( FthEMF+*@$33B%BoCOD@0$9=ARIAL6@intra_ordering_harnessC>""">>""">Ѽ>""">>""">>""">U?""">?""">#?""">/?""">h@?""">:P?""">rL[?""">3c?""">t?""">?""">?""">?""">{?""">r?""">r?""">U?""">9?""">??   % T DRAA PLxintra_ordering_harness% F@4EMF++@ @$foCwfD4C A$$==% % Ld9O !??% % $$AAFEMF+@<0Ne>@H<foC =mDXC =mDXCwfDfoCwfDfoC =mD@$$==_888% % V0P;KP;K99P;% % $$AA( F$EMF+*@$33B%BfoCwfD@0$9=ARIAL6@fix_verb_order ?""">9?""">P?""">%?""">Z5?""">C?""">UeS?""">9\?""">l?""">ǁ|?""">)?""">9?""">ϒ?""">?""">??   ( RpArialMonotype:Arial Regular:Vek%`h6`h`C`h hB`wTwwpww @ w `$wwwww@ 53@ -w w< !dv% T&mAA&Lhfix_verb_order% FEMF++@ @<0Ne>@4(}/VCAX CAX C$A@$$==_888% % W(Lc ((% % $$AA( F\PEMF+@<0C~սAX C&BC~սAC~սA@( $$=='%  % V,FN!|c||%  % $$AAFEMF+@<0Ne>@4(GCV6ByQDV6ByQDqOB@$$==_888% % W(,5IU!U!>% % $$AA( F\PEMF+@<0JDRJKByQDYlB6XDRJKBJDRJKB@$$==%  % V,2<!.U!!.!.%  % $$AAF|EMF+@<0Ne>@, yQDrByQDB@$$==_888% % W$UeU!dU!5% % $$AA( F\PEMF+@<0JDxByQDB6XDxBJDxB@$$==%  % V,bk!$U!!$!$%  % $$AAF|EMF+@<0Ne>@, yQDKCyQDC@$$==_888% % W$U![U!+ % % $$AA( F\PEMF+@<0JD5CyQD7C6XD5CJD5C@$$==%  % V,! U! ! ! %  % $$AAF|EMF+@<0Ne>@, X CaCX CC@$$==_888% % W$I L% % $$AA( F\PEMF+@<0C֎CX CWCC֎CC֎C@$$==%  % V,FN"c%  % $$AAFEMF+@<0Ne>@4(OCCX CCX CqC@$$==_888% % W(L 8% % $$AA( F\PEMF+@<0C9CX CaCC9CC9C@$$==%  % V,FN(c((%  % $$AAFEMF+@<0Ne>@4(GC&CyQD&CyQDcJC@$$==_888% % W(IU!U!% % $$AA( F\PEMF+@<0JDsCyQDC6XDsCJDsC@$$==%  % V,!U!^!!%  % $$AAF|EMF+@<0Ne>@, yQDs<DyQD$1D@$$==_888% % W$U! U!M!% % $$AA( F\PEMF+@<0JDDyQDlD6XDDJDD@$$==%  % V,!@, yQDe DyQDD@$$==_888% % W$6MU!s#U!$% % $$AA( F\PEMF+@<0JDzDyQD_D6XDzDJDzD@$$==%  % V,IS!$U!$%!$!$%  % $$AAFEMF+@<0Ne>@4(dC[DX C[DX CD@$$==_888% % W(I^|%%'% % $$AA( F\PEMF+@<0CUDX Ch DCUDCUD@$$==%  % V,FyN'(c''%  % $$AAF|EMF+@<0Ne>@, X C.'DX C}#,D@$$==_888% % W$IL) +% % $$AA( F\PEMF+@<0C+DX C-DC+DC+D@$$==%  % V,FN*~+c**%  % $$AAF|EMF+@<0Ne>@, X C4DX Co9D@$$==_888% % W$IL/-l.% % $$AA( F\PEMF+@<0Cl9DX C;DCl9DCl9D@$$==%  % V,FN\..c\.\.%  % $$AAFEMF+@<0Ne>@<0X CFBDX CED-BED-Ba;GD@$$==_888% % W,lL0k1k11% % $$AA( F\PEMF+@<0?BFD-B IDBFD?BFD@$$==%  % V,ir%1D211%  % $$AAFEMF+@<0Ne>@4(SCoLDX CoLDX CZND@$$==_888% % W(0L:1 333% % $$AA( F\PEMF+@<0CMDX CODCMDCMD@$$==%  % V,F7N@p33cp3p3%  % $$AAFEMF+@<0Ne>@<0X CVDX C^]DC^]DCʥdD@$$==_888% % W,IYL5X7X7*9% % $$AA( F\PEMF+@<0 CRcdDCwfD'CRcdD CRcdD@$$==%  % V,FO99e99%  % $$AAFEMF+@<0Ne>@<0C =mDCtDX CtDX CAvD@$$==_888% % W,ILP;===% % $$AA( F\PEMF+@<0CtvDX CxDCtvDCtvD@$$==%  % V,FN=#>c==%  % $$AAFEMF+@<0Ne>@<0X CNDX C> D-B> D-BJD@$$==_888% % W,lLO?AAD% % $$AA( F\PEMF+@<0?BD-BDBD?BD@$$==%  % V,iLrUDPEDD%  % $$AAFEMF+@th %CTDy9DTDD !DD}%Dx9DJ(Dy9DJ(D%CJ(DWiC}%DWiC !D#CTD@C^( $$=='^C%  ;'6O$'X($z($m)O$9*Y$O$9*9*X(m)z('=<>zL % $$AAFEMF+@<0Ne>@%CTDy9DTDD !DD}%Dx9DJ(Dy9DJ(D%CJ(DWiC}%DWiC !D#CTDx9DTDD !DD}%Dx9DI(D@$$==_888% % ;'6O$'X($z($m)O$9*Y$O$9*9*X(m)z('O$'X(#z(#m)O$9*<@yM% % $$AA( FEMF+*@$33B%B%CTD@0$9=ARIAL6@ Extrap.xml9N >""b>1W>""b> >""b>ؗ>""b>ʪ>""b>9n>""b>>""b>>""b>( ?""b>"?""b>??   % T1AA L`Extrap.xml % FEMF++@ @<0Ne>@, D C#D4C#D@$$==_888% % W$((% % $$AA( F\PEMF+@<0C$DXC#DC"DC$D@( $$=='%  % V,6)((6)%  % $$AAF|EMF+@<0Ne>@, D C5SD4C5SD@$$==_888% % W$KN44% % $$AA( F\PEMF+@<0C?TDXC5SDC+RDC?TD@$$==%  % V,HQ5445%  % $$AAFEMF+@4CAKD4CZD4C (\DID^]DwMD^]DQD^]DD (\DDZDDZDDZDDZDDZDDAKDD1CJDQD IDyMD IDID ID4C1CJD4C@KD@f( $$=='f%  ;266X@  7 X7!X7#X7% 7%6%6%6%6Y$%6%2X4%2#D2!D2 D222=<>$Qv % $$AAF0$EMF+@<0Ne>@14CAKD4CZD4C (\DID^]DwMD^]DQD^]DD (\DDZDDZDDZDDZDDZDDAKDD1CJDQD IDyMD IDID ID4C1CJD4C@KD4C@KD4COAMDIDwNDwMDwNDQDwNDDPAMDDAKDDAKDD@KDD@KD4C MD4CNDIDODwMDODQDODDNDD MDD MDD MDD MD4CwND4CODIDm-QDwMDm-QDQDm-QDDODDwNDDwNDDwNDDwND@$$==_888% % ;266X@  7 X7!X7#X7% 7%6%6%6%6Y$%6%2X4%2#D2!D2 D2222X@ Q3 3!3#3%Q3%2%2%2%2H3X@ 3 3!3#3%3%H3%H3%H3%H33X@ 3 L4!L4#L4%3%3%3%3%3<@#Rw% % $$AA( FEMF+*@$33B%B4Cn-QD@0$9=ARIAL6@ Order modelUC>DD>*>DD>r>DD>>DD>c>DD>UU>DD>Ǒ?DD>C?DD>U)?DD>8?DD>H?DD>??   % TU:cAAa LdOrder model  % FEMF++@ @<0Ne>@4(dCVCX CVCX CjC@$$==_888% % W(Io o % % $$AA( F\PEMF+@<0CܳiCX CrCCܳiCCܳiC@( $$=='%  % V,FN!c%  % $$AAF|EMF+@<0Ne>@, yQD 5CyQDBC@$$==_888% % W$U!Q U!" % % $$AA( F\PEMF+@<0JDACyQD_IC6XDACJDAC@$$==%  % V,! U! ! ! %  % $$AAF4(EMF+ @$jC+C>C A( $$=='% % Ldz !??% % $$AAFEMF+@H<NnC@@?@H<jCICXCICXC+CjC+CjCIC@( $$==_888% '%  ;6X46X4=[6X46[X4NCCCN[=6X4,777,6X4=161X4"1-<-J-W"bb6bX4bWJ<11=v161X41<JWbb6vbX4ib^W^J^<i1v1=1691X4G1R<RJRWGb9b6bX4bWJ<11=161X41<JWbb6bX4bWJ<11=-16^1X4k1v<vJvWkb^b6-bX4 bWJ< 1-1=161X41<JWbb6bX4bWJ<11=R161X41<JWbb6RbX4Db9W9J9<D1R1=161X4"1-<-J-W"bb6bX4bWJ<11=v161X41<JWbb6vbX4ib^W^J^<i1v1= 1691X4G1R<RJRWGb9b6 bX4bWJ<1 1=161X41<JWbb6bX4bWJ<11=-16^1X4k1v<vJvWkb^b6-bX4 bWJ< 1-1=161X41 < J Wbb6bX4bWJ<11=R161X41<JWbb6RbX4Db9W9J9<D1R1=161X4"1-<-J-W"bb6bX4bWJ<11=v161X41<JWbb6vbX4ib^W^J^<i1v1= 1691X4G1R<RJRWGb9b6 bX4bWJ<1 1=161X41<JWbb6bX4bWJ<11=-16^1X4l1v<vJvWlb^b6-bX4 bWJ< 1-1=161X41 < J Wbb6bX4bWJ<11=R161X41<JWbb6RbX4Db:W:J:<D1R1=}6}X4}6X4 } }=}i6}8X4}*   *86iX4v}v}i=}6}X4}6X4}}=B6X46BX4OZZZOB=6X4qfffq6X4=6X46X4+555+=6ZX4LBBBLZ6X4=6X46X4=f65X4((56fX4ttf=6X46X4=B6X46BX4OZZZOB=6~X4qfffq~6X4=6X46X4*555*=6ZX4LAAALZ6X4=6X46X4=f65X4((56fX4s~~~sf=6X46X4=A6X46AX4OZZZOA=6~X4qfffq~6X4=6X46X4*555*=6ZX4LAAALZ6X4=6X46X4=f65X4((56fX4s~~~sf=6X46X4=<?v % % $$AA( FEMF+*@$33B%BjC+C@0$9=ARIAL6@ switch_heads?""">r"?""">r8?""">@?""">r J?""">9Y?""">9>k?""">{?""">8?"""> ?"""> ?""">ǹ?""">??   ( RpArial%Monotype:Arial Bold:Versik%`h6`h`C`h hB`wTwwpww @ w `$wwwww@ 53@ -w w< !dv% T%mAA% Ldswitch_heads % FEMF++@ @th %C.ZCy9D.ZCDdzCDRCx9DCy9DC%CCWiCRCWiCdzC#C.ZC@C^( $$=='^C%  ;,6O$,X($$O$Y$O$X(,=<>rL % $$AAFEMF+@<0Ne>@%C.ZCy9D.ZCDdzCDRCx9DCy9DC%CCWiCRCWiCdzC#C.ZCx9D.ZCDdzCDRCx9DC@$$==_888% % ;,6O$,X($$O$Y$O$X(,O$,X(##O$<@qM% % $$AA( FEMF+*@$33B%B%C.ZC@0$9=ARIAL6@ switch.xml>""b>9J>""b>r\>""b>>""b>9Ϊ>""b>@>""b>>""b>DZ>""b>Ǒ ?""b>C!?""b>??   ( RpArialMonotype:Arial Regular:Vek%`h6`h`C`h hB`wTwwpww @ w `$wwwww@ 53@ -w w< !dv% T1AA L`switch.xml % FEMF++@ @<0Ne>@, uC$C4C$C@$$==_888% % W$qq% % $$AA( F\PEMF+@<0CCC$CCcoCCC@( $$=='%  % V,q.%  % $$AAF|EMF+@<0Ne>@, X CIBCX CC@$$==_888% % W$I;Lt$% % $$AA( F\PEMF+@<0CCX C+CCCCC@$$==%  % V,FqNzc%  % $$AAFEMF+@<0Ne>@<0X CICX CCxBCxB~2C@$$==_888% % W,iLJ""% % $$AA( F\PEMF+@<0BBCxBCBCBBC@$$==%  % V,fovfvv%  % $$AAF4(EMF+ @$oCxD4C A( $$=='% % Ld#>O !??% % $$AAFEMF+@<0Ne>@H<oCNDGCNDGCxDoCxDoCND@$$==_888% % V0?I?I#>#>?% % $$AA( FL@EMF+*@$33B%BoCxD@0$9=ARIAL6@assemble_compoundsc>""">>""">x>""">U?""">?""">x)?""">J9?""">r??""">9nO?""">@_?""">xm?""">J}?""">9~?""">g?""">P?""">8?""">!?"""> ?""">??   % TAALpassemble_compounds  % FEMF++@ LdrTqS)??" FEMF+@ Visio (TM) Drawing TRl !fffMMM333^CfU8@ TZ Arial@NdZWingdzs@vmuMonotype Sort NSymbol5T?? Y@-1U J:DT1EW-hTT<* /Ub bO0zGz?@8@H2!kWb*Uk9 +HPL/^&9^$?Ak^&*,,'%/v&&U  1y   )? 2    12?aBBHEHEHEHEHEHEH@?>?:`T2BBHEHEHEUHEHEHEHE%H@%O9 F7AOY@;P AsVsVA!gLTkY W_W__ !`#k4lb6u`kW *4l 4l %Y?P:?-\ #!+|QtKf2|2|2|I2wGQAUoTMeE$ttA%_8BOTOfOxOO??HO?7ܻXuW?YsU42 T*?P?޾B~;$^ ????+P?OO+O=OOOaLqOOOOOOOO__%_7_]?Gh_Ռ249_ﳸ____ oQ%+o}RodovooooooooI_[VK}??? 1ASewяR#ЦUf/ ҟJ/\/n/1CUgyӯ -?Qcuڿ$\yԁ1λ|Z|(g${iPхϠA32O?贁Nk@RFRR#Y^Ç?(:iӿٟ` S odXXLetterO_b PRIV@|rpE fg4o/#/5/G/Y/k/}//////// ??1?C;} DisplayB5A7UFDfP h-RTUUUA@ ?I? 3h  ePqYk  KHo:Lw >2QO` Flowchartw`u D(2OVhz~ WThis symbol represents any kind of processing function. Double-click to add sub-page.H D  # =hj0>TdYY9 ͉UA@ ??P6 u` 6u e.A&a8 e&0LgH>v5 LS{5 `7Copyright 1999 Visio Corporation. All s reserved.`_BFS.chm!#22268D" 0d9 l>YUdv "}T <b;2F6& :2ge24 7=H?{1>%??K91{1^;.6=G=E%2O;rA wB}C! ai;(!aOj_(O_k,OqV(@^0A^ 5F@dL?VVG~#A&QR?_\5C _T oo!TS>%[fojoQ P3m5#_d^*+T$]tjt>"e}sIsQ q?B`Cost6,Enter the cp associated withpis process S"`@ zq`vQ:bsDu* wd@ {ofsstep @zq>%$s Resourp:wnumbpdpeop{le quirpktopm܀testask n>!auY=@|v=# %Properti:0Se}tsustomqsce sele}cqshape s{ wBb  f ,f2^*mb?ؿ贁N8i6??54 _ Q);M_>m w =_H'-D !OyaGE_5>FYw#C!MB t"j]fa]@Zk]P+84aTd9pE8= I|* UFDf h-TUU[U@@??I?`d buoqYkQhu23u` Connector ` e1Crw UH ^   -}|5 p`I`V?}`xa833ސ]3σ3u 33?, ?Gxpx^& CThis connector automatically routes between the shapi ts.b?2jZ/0? HD @# =h8T YY9 BT#F oU@? P6 u `u bA@]u  .(#DB uu`h?\hr|uVa@-?bl;'bE-ho'$y( 2rq?@I ?$%? @"U*5L -br  ^vv"(2uI."q28v"uh9Bd&</MS{ #145 `Vis_BFS.chm!#22291`7Copyright 1999 @io Corporation. All $Ud vE \/4 *&1$b24R(^[w[D ZQi a59 93O'2"q?/g;2GHu-!N !OyaGE_L>F=`w#ߔC.7B uA` uVM J.AJ<-QJ*(.88YFxdnJ>ʷI@>brM J?V,"rZ|-#%))<"<>h>@/<@SNG1L4+Ir"`?Copyright (c) 2001 Microsoft Corporation. All q2s reserved.l`Vis_SFB.chm!#265119S3!tA7 30lJ0JUhh5!*q+#`#5brn"u$b@$ `UD"B($A'3`#N89 _EA$^EA85# OG>:%?RjK_ lP\RV:#{^( bPA,b(EGOO@a`Y?ffx3MIaofBDa(2v#b%d@oA}a5Yi'Z yI@Oi85k {$_gAQA(D<5oicOUwKHa^)y@Q%8_[?n[RVdMle85ki\f"\#B2#FA@CA -@B-J;#mQhe$67+B>T11bR"o2sN G5G1a b`Cost6f3,Enter the c associated withѠis procgess0 B`) %`a#Du3*f3 ̧dE {ofstep0D5{$( Resouri0:̧numbϠipeop{le0quirktoՠmte/task0nZ:LJA7O1G1I3 %Pr?operti0Rb ł01Bce>՟I5z{~!g)dP(:hTA?"aa2(QG1e`Dct,data,dLly,acible,such,sto,onVsk,drives,Basic,Flowchart,inoform2,f~Vagram,joiners,ۥ0!5S $@=H-H? !OyaGE=>FR4#XVB{ CY]RaDS@+,/Z5a5=PUFDf h-TUUU?@ ?6I?d XboqYk*Qu23u` Connector} ` e1Cw \UH ^   -}|5 p`I`V?}`xa833ސ]3σ3u 33?, ?Gxpx^& _Connector that automatically routes betweene shapit cs, using a right-angled line.b?2jZ/0? HD # =hj8>TA YY9 T#FAoU@? P6 u `u bA@]u  .(#DA@uu `h?\hr|uVa>b@-?bDl;'IbE-ho'y( 2rq?@I ?$v%? @j"*5LA-brB  ^vv"m(2u."q28v&"uh9d&<?/M) #145 `Vis_SFB.chm!#26514I`?Copyright (c) 2001 Microsoft Corporation. All $>U@dv5 !\^4 *L1$bC24R(f[[D ZQiAa59 93O'2"q?7g;2GB81aj e/}@~!gdao7 T KaKa 2(c1B1@Dynamic,connector,autom[@cally,routes,betweenus,Fy%w#Z&7B ~'gdo@{+ "st(PUFDfP h-RTUUUA@ ?I? 3h EePqYk  BHo1Cw Q2MQ` Flowchartn_` _ QJ /?|wxpf6lbx{s؅pwp  wwRIndicates a file composed of records thatntainelandset operations.UH TD  # Uh8JT eeYRJ>UEMUA@ ??QNuQ`t?u,AMA JD$9A 9fQjQ   k@QM.B:u\zaaJN[8!@P `rpr{|" Nr\wb#$u8H N# "HoLJ#)1.4+N`Vis_SE.chm!#29529I"`?Copyright (c) 2001 Microsoft Corporation. All n2s reserved.S>#7>K00RQ) #53sV##0O贏Nk?HH&QR ##66H #BCMlJY0JUppY5I q! (w*+ 8"?4$bS"$"'.aA aM  " k(WMdQmO*$ VHSf8*cU/Q&t@ IzW$"d$  !%SuuXf FQX3]OlSQ5Oc [?_}Rc#mm 2$|Srp[AQkDH5I OUKSoowZy S26WM Q%zomcShoBf u%d%7} RW 5XdM!mX) % H&bVt$@bT2rq'C#$b"1w#'0YвXXg !V i!VӓٓW8 j[AUOPas~a}3)ph pH%| | ] og { P`/v,8|Nj Mc[AYK]o~!V_6 p"ϯ+{ R%_37%CUYh r){&+8tgcqA(j5 IfpH&H&SE`nC-!V|т %O}߽߫y\P|%V/#/%DB|5c |anbߏJ#711)1CB` %Properties{c `0}Rb "0@m@0Apm7ɱ+NT!! "Ρ΢*a*a2 (s@t)5)1.3Ct'Cost6c3,Enter the cn associated withis procgess02`@.%=S.#`Du3*c3 {d {ofstep0.5{ (` ResourDc3:{numb~people0quir֒tomte^task0ix0BDtK'2q BBBǢ$Fӵ8KcKgTtAA"!!2 T{.1K0Database,Indics,file,osed,records ntain el set,2s,Miscellaneous,Flowchart!nectors,links,joi0C1e~:0inform2,sy ms,d?iagram!X+c9$@"mb?ؿ贁N8i6??54 _Hu# # !OyaHEhFĒ,P-B  .YaDYfo@tkgoiP+d54aUt4> ףp@bq?$@ 6}C-7"A8t4> ףp@bq?$@ 6A~-78U=?@t4> ףp@bq?$@ OK7=C-b7"A?@It4> ףp@bq?$@ T7A-7;U?@|t4> ףp@bq?$@  8C- "7A @$6IR@7DR@$d7HR@7KR@l(8KRH<( H<( H<( H<( H<(  EU9 RE$b9 REo9 RE|9 RE9 R_f(2T ݨS6*BV!:/{~"DR$t"v+w6?l9s8wPD59* AM1AM1PPTPPT1h/l1; A/>; A/K; A/X;AV7V@U=??g @9OasN(m*|; MCU p@Ibq?$@2m?I?.CUg` 46 2@?qu` ""&"u#)+ ?A/-'q/sB#&t/oY%*QF1F1 ` Flowchart `g2 m:31@??Q?c9Connectorx?:eU!0D aVj  4 1d0 T !4!"#%&T*+,U-./D1%235g1!1HA@DWDaWD>AW>AHBAWBAHFAWFAH1W5DNAWNAH1W1HVAWVAHWD^AW^AHbAWbAHfAWfAH !W !HnAWnAHrAWrAHg1Wg1HvAWvAHWzAHWHAWAHAWAHAWAHAWAHAWAHAWAHAW$%DWAHAWAHAWAH'ZHWAH)ZAHAWDAWAHAWAHWAHAWAH7ZAHWDAHZ"AH;ZAHAWAHWDAH8ZH9ZTAH:VGPA!%ADAAAaADA>AAABADFAAA1ADANA1DVAADA^AAAbAAAfAAAjBAA"nAAArA"A"A"g1AAvAAA"77ƂAy𓱆A}𠱊A..1N𺱎AXDZƒAOԱ[ᱞA\Ac]A^_A20"Aa/Ƅb<cIVAccApk00A A@0cXA0Q*31B@@$?jbAY%AII. 0vert to Basic TreelDr߿6IKኲvAV( 6E7@~ fUFqA4eELw', Dp:LAllvgY%$,Uiئ4DH'2Wq `mCD! 14/]^EATAADqqAA,RQQ;0d\n߀ߒ> ߣp @#60- 2SSFDB^gzF|w #5l_^m5Њ|b_t_harness? .3oG\ao7oo"wC/XoSAzOw"Fgb{"F8/oHoHX/j/|//////// ??07c1>?P4E_?q????_???]M J#O5OGOYJjjCCgҦ|OOOOO` __-_?_Q_Tu__|)1introduce_coordination__ oo$ooHo]vooϾooo*<]r|^@ugpdۧۨSs3lR?i羅ȏڏ"4FXj.9ጟnџ=+=OVhz@jC$O 2DVhz /Ϳ YNupdate_parent_-[Qcu#/L?;M_ߗ/u??5O>OoOO(s_,>Pbt|`d#5G/k}|Gz?`2!nPv:dvdq<PqN`rw/DFQTi/#/Iaa=)QIII. Global Movemb f/x/o///n?/g?M6?#Z?~??????-O@OLKfA_YwYxg㪯CK vwO`SOzOdd@_R_d_v________o(oEoio{ooӏoo 6 1C Ħ"nâ<ц(:Ŧi=)ġg_mvmt_harness_pullȏ4sIbwΟ(}I^N߼֯\X>ؾؒȿڿ"4FXgx:dřϽ+)bBTfxߊ2Lq%/V6&F1|sA 0BTfx5O%1sh J?@Rd_Eo*<N_rooy{J9O*st?v //,/>/P/b/t////h //N ??1?{_U?g?y?7L֋/ @?????'"Ha\Ha*OOPO apOD aO寱OOOO __:_L_^_p__HFF%GEx___oo(o:oLo^opooХϱoQi elat?ivizeroo,F4FX|# 0BTcv%Xڏ-P^p@џ+=OasbGGԯ5 b@Rd|G?@L֋/6¿Xv`3/& )GC&vSC!'9K]o|uƉϦ͡T@ 5?6i'kIV. Intra ConstituB OrderingEW߅{ߟMO6O O9~O]og@*_L'L(@_Xk<__B3-@Mo1CUgyƧ]q$HZl?d> ףp @//۱VV*GR Ýe/w/////ϓ/??`zuH?u]q٦&i1_o4_harness?????OOZSO wOOOدOOO#__(o=_^;_ʿ(ϿSo_\{oooooooo /Aa~ߢ@??ˊmT2DVh ,,hf[#.V῏я+=OO贁Nka5uTf9͛assemble_compound0&8Jnȯگ%"4pXNѿֿ_//Uv/T=/xa/P_e/:L^p߂ߔTvא1?߀\nM@@Ӯ+$UD@gF̤yFJ9du `u AIp1զZ괩¤KʠxrĢ $@pňjwP@z3dj|pu1ξq‘Ü5XFJ@yE2r)?@IW?MUſK(&OXPr^QEj_|P1ሑaχ P4Q58SSQP5 tNqzsB _ҸqhY<4o b,Yaq&AskbWD1oUoh~:ӦN3ϪO1SV>PbtQF[#UDԭT 0f~FufFA#aqҿ@'wQhTg ŽQɹ(tCUgyϋGϠ);MtnwQ}ߏԎQD/g-Q,>P ~u ? O1? 1dCwfo_"UE ܥ%""F O\" awS2}OO VO_"^hu;H ._Rdv_l5k6;f;f$'(:Q(/O1o W/mO1}%p1$':Q)%-//j_ \/j̝P?b7qp?????0??O!!O3OEOWOiO{OiǠ @!a:QE&_. o[nK_]_____&___o$u-vbotoog(oԭq+oNq*<Nrr#!|#š'%A*A)KšQi"> opMk vŀu`u@ ^m ,~ׄ~ߐb!ֽ߁ߢ " "BK]o?cQScү䯨.!47 呅ߟ@>T8Jd/*៿/&޳0?p???O OX2ODN¶\E nMPOOOOL~ EGU>_'TŸRRX_R>YS_ ɴF__ ooد@oRodovo֯d[ ZLRooo|x+>޿F>l~}3#?Q_߃ˏݏ6eD+4bbjDum0`uoPSpWG+ZZlQSS4S ףpM"7M%"M%4F// ^b '5Akd\wB?\nA/vg?NQW ?M%????O)O;OMOFef/OOOGEOԴQ_ΝQ);Mφ__ͺ_Ut`Þ_ o.o@oRodovoo.C{;/M/_/q/%C|CvL?֋/'E('9ϩs`vzfy➁:??pP߸ʏ܏NӥҦ@RdIVyIB_T_f_x_0 _f_d'|6ׯh 1CUg Pп 'UŨsuy269K -?cu߇ߙpD SRZCD&6瀘:D6/i-6 R!:h!W?{oa-.` `f+oL^%yIp|G? AU*V)_;_"sx|Vh_{_4r%\V)WeOIso//qo56R1fk/}//'Q/bqx/}bqXA,7IieN?`?K?EgFXh)? d?? OO1OCOUOyBįv #OI<O+} QDU$_1tr^q_R`Yço 5+o=oOoaoHooooF˅Ӽ¢,S|+蛡\LNϋq͡5ׁ84tuA3`/*YϏ߀);M_߅Q P]M 0jZ/6@,>Oڟt Pt?懯אָFYO@_'=]>za/߰ߠ_ eE6L?KڣG 4Oaٔtɶ?@?EˊmT_kLT?Q>Y7uTYSF?Ow@@T~ʬO*<OdL%K&Bc=VMxnt>_ ]>2DVhzmKmof#mf 54/~SP/7t//////RfPk@@ _ g?q4` t٤G?Y?ć6ڗڨ?@| ?@iWz? ǩxx%4幆(OMO_O@qgOOOOpOO_Y5di$ "uM!ve_wS~_$Ƕï_@ȶk_o&j^s]oρl vk ʯo6uSpqLwmtQ*Իdu0qwwqq 逸qQlq)["\Zت|LQ`şݓ瓕0U@Բ$|9َĆsU^C\bb^?bCzQu ɖQ mtwqttc7U+T*7`dT{qttLy%8e¨cqö1oCo~(%{] aqQ\vvϥ>git,QQ11`U/pqy~?ԿpqaݑU*cu,P!vrͯG_0d  #å,(.J/`(*.؅/T,e(+{/)qsSxtrap.xml rʱ |`ƿ<Nwϥ`ytt u@G O^_rVl 5 5 5!L 22F3͘\^jc> jǰpvFΏONr|W ˟ CcQ:0pqOuAI);M.hho;Q#@S0O贁Nk)$``YcQo~mcQ\GC&nw0Fq -(?QZ16iJ֋/(E 5 5 5?Q,c_7VPOǬ?_ ӟQϥՇ4XG{2bqa?ϟbsd?Ѕx0j[UwfzbfbPѣU[ ̇y&q\į"4b 46G^aė7]^+?//bu˿ݿP%:#c/m/qu1ɯۯ&;jq88Order modeli%4QS7~|@FFdAdOcOBYI __1_7oUSYXs___RY^d_U!__ oo/oSoeowooiYloYoo+dOay^ϯv9?73E4 $U@> ףp wąF ҿduL#`u$k߿@qW@DA"y!3零pTAt`sVR Bϟ._S-,iR;MSXKXS[شƯ دVTPs ɿۿ`kQ!ϻF Xj|ώ}:+=߰O ĄԠ@֦ WÀ|Gz?U6??COZ6F?Y?BI<67'03F2_S3FUg_S1Z)_(:TV_cy],-Y _%cjm'6eBon(@*<N`r};a"/!/3/Vo1@@Ӯ!@w%׉. /xpKp{ @ mL`o?_q??_Haca5 [OmCtOGtaOOOO_ݏ9_aai_5a _fo7do̅oooooUU!"#U$%&'U()*+U,-./U0125U89?@UINOXU[\]^U_`abUcdfhUorsyU|}t45> ףp@bq?$@ 6^C-tu A@\uLRH<( Eu R\v<7Mt@? `u.PDu.PU1( UO"D&aU=QJf )h"TyqU-xF 5+7 8 : Ʌ&Q- H*9(T*EQ6/H/Z,GuideTheDocPage-1Gesture FormatVisio 90ConnectorVisio 00Visio 01Visio 02Visio 03Visio 10Visio 11Visio 12Visio 13Visio 20Visio 21Visio 22Visio 23Visio 50Visio 51Visio 52Visio 53Visio 70Visio 80Flow NormalHairlineFlow connector textProcessCostDurationResourcesProcess.4Process.5Process.6Process.7Process.8Process.9Process.10Process.11Process.12Process.13Process.14Process.15Process.16Process.17Process.18Process.19Process.20Process.21Process.2Dynamic conn?ectorDynamic connector.25Border GraduatedDirect datavisKeywordsvisVersionBorderDynamic connector.8!Border Text Transparent LeftProcess.56Direct data.57Dynamic connector.30Dynamic connector.33Dynamic connector.34Dynamic connector.35Dynamic connector.36Dynamic connector.37Dynamic connector.38Dynamic connector.39Dynamic connector.40Dynamic connector.41Dynamic connector.55Dynamic connector.57Dynamic connector.59Process.1Direct data.68Dynamic connector.69Dynamic connector.26DatabaseDynamic connector.27Dynamic connector.42Dynamic connector.44Dynamic connector.46Dynamic connector.56N 3w E3w E3w E3$wG3?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~UnU U UUU;CLt4> ףp@bq?$@ xXC-  7A%t4 _" A-3,7AJ@։LR@<.5RH<( H<( JE RE$ R{ : g'9K] (h(|+@(?3?I9" [La:]BU }P  bU\#wW&#N!co>`#&F;39F| xR$;wP.KfxN̳ޥ)_K<+'duh11vp=Cħ2<~b T?c*PD6%?! `oь C{k @VisioInformation"OSummaryInformation(DocumentSummaryInformation8P$_1081080974!FPL G0G      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnpxstuvwxyz{|}~Oh+'0@HXdp|mgamonGm  EMFl@VISIODrawingL,mn ??d((nﯯiii000````````````````````````333|||```````````````xxx XXX qqqqqqﯯNNN```翿翿翿翿翿QQQ翿翿翿翿hhh@@@@@@@@@@@@@@@PPP@@@@@@@@@@@@@@@``` XXXXXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@ ǿǿǿǿǿQQQǿǿǿǿKKK```翿翿翿翿翿QQQ翿翿翿翿hhh@@@@@@@@@@@@@@@]]]@@@@@@@@@@@@@@@PPP XXXXXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@(((ǿǿǿǿǿKKKwwwǿǿǿǿ``````000OOO@@@___www   χJPX06@ !'006@tvx///!'002xCL`ffffffSL6 .` ```^bh<ffffffffffffffff0```TY`@ffffffffffffffff  000OOOTY`@ff@ffffffff  @@@kkk{{{TY`@fff3f3f3f3f3fYff   ?DK@ffffffffffffffff   DDD:::@@@@@@@@@@@@@@@@@@@@@@@@@@@$@ffffffffffffffff   TY`@S6#X@0@0x@\   TY`GGGTY`0@@@@@@___ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@TY` *hI\ffffffffffP@ P___06Lfffffffffff\F /h9CP@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@>>>@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ tvx06@#@#@#@#@#@#@#@0AGPϿ333000翿翿翿翿翿QQQ翿翿翿翿hhh@@@@@@@@@@@@@@@XXX@@@@@@@@@@@@@@@``` XXXXXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@ ǿǿǿǿǿQQQwwwǿǿǿǿ``````翿翿翿翿翿QQQ翿翿翿翿hhh@@@@@@@@@@@@@@@eee@@@@@@@@@@@@@@@PPP  (((III``````߯BBB!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y)kppp@@@ 'MC^C^C^C^C^C^C^C^C^C^C^C^C^(85,.7 ___!/yC^C^C^2FC^2FC^2FC^2FC^=UC^ ,r ,r >>>888@@@@@@@@@@@@@@@@@@@@@@@@@@@ )jC^C^ C^C^C^C^C^C^&c(8 #.kC^C^C^2FC^2FC^2FC^2FC^=UC^!/y#[  000'C^C^C^C^C^C^C^C^C^C^C^C^C^4IabhPPP78?``````pppXXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@߿```ǿǿǿǿǿǿǿǿǿǿhhh```PPPwww(((@@@￿￿￿￿￿￿￿￿￿￿ϟ```@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@pppǧ ```XXX@@@XXX@@@XXX@@@XXX@@@XXX:::+++XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@ׯǿǿǿǿǿǿǿǿǿǿ```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@￿￿￿￿￿￿￿￿￿￿ϟ```@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp}}}@@@ppp@@@ppp@@@ppp@@@ppp@@@pppϷ 000 ??? ```@@@@@@@@@@@@@@@@@@@@@@@@ϯ@@@@@@@@@@@@@@@@@@@@@@@@@@@ϯ߯```ppp@@@kkk @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@  000PPP @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@׿ϯϯϯϯϯϯϯϯϯϯϯϯϯϯϯϯϯ000ϯϯϯϯϯ222@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@߿￿￿￿￿￿ZZZwww￿￿￿￿￿">">">">">">">">">">">">">6HHH@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp{{{666ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@">C^C^C^C^C^C^C^C^C^C^C^C^C^.A EGO#.kC^C^C^C^C^C^C^C^C^C^C^C^C^!/y&chhh+nC^C^C^C^C^C^C^*;C^#[*;```www"/rC^C^*;C^!/yC^!/yC^!/yC^!/yC^C^ ,r ,r hhh@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ppp$EC^C^C^C^C^C^C^C^C^C^C^C^C^,> <>G翿翿翿翿翿LLL翿翿翿翿翿!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y!/y ,rKKK000OOO000___www```````````` 444@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@***@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@```HHH`````````HHHHHH@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@...:::@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@pppIII``` ``````@@@@@@@@@@@@@@@@@@@@@```@@@@@@@@@@@@@@@@@@@@@000@@@dddNNNǿǿǿǿǿǿǿǿǿǿ```XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@ppp```@@@```000000@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp@@@pppppp￿￿￿￿￿￿￿￿￿ ooogggPPP`````` `````````000!!! ```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@::: @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@xxx``````@@@@@@@@@```xxx```ppp@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ FFF ```@@@@@@@@@@@@@@@@@@@@@@@@ϯppp@@@@@@@@@@@@@@@@@@@@@ϯ```@@@PPP888@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@   GGG  @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@՜.+,D՜.+,h$@HP\p |   Microsoft PagesMastersPage-1ProcessDynamic connector Direct dataDynamic connector.8 Database8_VPID_ALTERNATENAMES_VPID_PREVIEWS_PID_LINKBASE A  FMicrosoft Visio DrawingVISIO 6.0 ShapesVisio.Drawing.69q ՜.+,D՜.+,h$@HP\p | z; ߁_46/ _ $>s͡CB.b:YɬL_L πππ1_| ztbW[BKk.4jG3'UTuE /Af%/vQNy28/34j /%p_ėk?_ |mh&]Dv_`K÷=41 4\' J3 ?_Y%[ _| Vsw%K·'4B94rhBs_jC_ fa/b/2$14>>>: q%4щ/_=l /NkBʹ nTO_|wuO?Q4F:I8 ZtƼ/ܗ!%5O/_I>ݐj>^'|u·{U+Ww1q4z A>*A~Og|! 7TnדWWf_c̳'s}+;aG#U:n{ȋUGOxsa{ v.䳗UɃk|EȃviAt r&ypz "rnQ3j.ȃ쬵 Zw<7K[Ddj+15ۮ:zU 9nj~SraG*ϐf-h܂-G?Q[Q*+r޶>NW|u+T|/c~ ߏa{[w|; iNQOUR`ZXq#C+ YQ U;+LO|W{篔 W W5OVϯ~g|×i>H k[hAk#[|3>Q6b<fjcʸ1W_ |ð]| ~o~ T? vǘ'2__ͷ,S?jVn3+4T͡qᛃɗw4F)W U|ϲ6Fh]DvC|"<f L M_#qшՈLo\o^Y _J+j__>؊.|"!e_<|yz% j -mAhn OAb2 11)0|%J·t_?|E뇭 MПam Z;cK5/!f_BK _%/78|s4C.+iS _$5zϲ6F4.|"Q/a5_G;+~Bլ3/a ?8q\v߸߼/%DW룿_|}]Dv_"_<|yzF$6V4Zڄ斈8DDŷ&,K֬\eVQ_4+W0leQhW_~ T fȠͽQbi,,_,bV/?|AӴ\WA _2|#WIkПem,Z=h]D~+ wPn_qʺyu`ºkg Us}'zrE WgU|ߨ?e}~:KM?a=[/W-o9|[j+ߟWݠ&kj*?W}Cch}~z<-y[Uy[ꛭuTF>EyzU>9yr˃|Q5߱-GDhC`tOh333ૃm,:/Khգ/_ z_ZoYh@3'2_.{(,"DCf*vQNy22ghբ3H]G%k/Ckk?_ |mh6Vm3 n O M Y; Z'<s<mυUh/| -w  %K—ol)Oh*ksh:恐15d&b/&b2+SS_333ૃmt_B|1bÖ|w .LF|Af_K0QdVA_B(kGE4'3>C_ eH|I 6SK׆f҅Od7%/f_4|* K@_`h@h_|"?/~ŏ_ %Ko%l1~hm/ _$|`KI|BS/XC#!4]Ũ04`b/b0+CC_333ૃmp_B|1bÖ|w .LF4,|(Ǭ|4Q֎1i4O\ghբ3_ /F |)LU:넯}ކn/^5qX;ZH7_|X3"3~%yϐ ֌ZO>YKKvM~V=wySMvTH{O'W}ҾjٿƾOٟ7KKtg-V>k͕?zg)O*g_{>ao?Zlٺ5_|q !3",u_W[+}>ʕ̙LeM%gi]k<ц J?$dPVɠ5( wЧ/_<|M MF[XۂF ZWv oqͪu UWōi0|%J·Ǩ+h7|CVo'fȠ=L>Zj~f῁ںjYo74ML7N2|eFė;LY4hF3'"2_&`vU`Oh鏳6FF4f||" dV_Q| Vty/ _6(KP_oam -hmDsK_|"YɬL_4+W0le=_4+W_?lOh* k3hdډ_B|!  1_BKU!_BB8|7M3/:eJ_^,khdA3'"UVuY7|u3uUGY% k qшՈ溰85EE/%¬$_ 냭'"U _6"KDͷVֶъ&4D\'2(*5Qf_Kt5QfEo|%J Vvu_?|E뇭 MПam Z;u/X_,_,feIbi M74M_|% t+7}$_YfȢՃfޅOd&jWUg(nTe(x:8B&؟n3T ߪ~NQEV+ݍ|*K&*U|]0FÍEݍn|m\ ,5 *Q "E n J_*lGC 7ZRA9ؠʠlȱ6FCh2(2`Tw,p4Z5lN M_||||uEg| ztbW[BKk7kwh]D^DE|EE@=Bjx(¬"|BSwQ֎1i4OFf5; ZtPC-eH||mK f>jaf_4||BSwA1 4g|"{.,BPs!fi/_|+a9Eo|I%[JJ?9y /_L 1ʔ `3]G'_ |%$_;]B v4.|"/R/A5_wY <dVA_B(kGE4'3>C_ eH|I 6SK׆f҅OE%/f_4|* K@_`h@h_|"?/~ŏ_ %Ko%l1~hm/ _$|`KI|BS/XC#!4]Ũ04`b/b0+CC_333ૃmp_B|1bÖ|w .LH4,| 1+/> hu͓>q|y/_/CKjџ/_ 64.|"/"u{|}N:o`CW8@chNePg%5ID6 xdPr$v_3'{;crŏ&Yn_[U@:;km&{v3HyCggL>[k_$;co{sev֤<*dKAO|o϶ϑ-干%Jǹ:2Kʆ++?[Ȑ_:&\ɐW2$sfdHD~䰔!q!?>r3?k&"S[1YL~ȯa*uY1?wIfZۏ^7b+~8=tt =Wl]#_}Xƽ}*;;s{>גqmoyzɫ*s;oi缏2yt{4w_k1s3SV=‿O[jq7;cs O~zC<z׍.ϋRKo~v0ӎ#<}kCaS쥎G'_{1?x'uĺ:ĺb;:O7neos_kع-[y؟ssu֗:ޣY7ͪ\w3?Iڊc=u̧nmǛoФʿM9}T'hooR߲}zGor筹U|w^v\M:z̷r_a4h/?|g|@snj~l_ Q rA쵦GŨ|W^V^Vbm@(>C[~P;S—/_^o#-mAhn+;D~PŷژvbTk4_Yo|%J VvcT4o!a+H|BSŷ k3hdډc&|-~5f?հvYo74ML7N2|eFė;LY4hF3'~2_&` 04T5gm8Zh +:5E /A%Ȭ$_ 냭'~2_6/×lP򗠆4Zڈ斠8D~P/&01Yh W_ a.|{LhW_~ T fȠ=ZCb!%k[C4qof_|% t+7}$_YfȢՃfޅOd%^7bTڎ0 K㬍Guaqk͋0_"KYI|E > +[хOd%/×lD򗈚omEMhnOU|k"Dh5QfEo|%J Vvu_?|E뇭 MПam Z;u/X_,_,feIbi M74M_|% t+7}$_YfȢՃfޅOݞ:vrUm姬\>B&؟].ߞ\'֞F4Z//_܍ _oBu`ªBɷ6km=u+ߧڛT|inn-M7|jR}8"ESMn445[Rۨ-M7Z75(9?xS*?xSS94rhBjcF略Q[EojN M_||||uEg| ztbW[BKk7kwh]DOE|EE^jL MiGY;(Z<34j Km4r_׬F |)lvY?+fK÷=|SSWxOhN?A4:LOU|υE^j M4V—/_\$|Im-% MПcmZ<r/_L_LfeJbj ||||u.Khգ/_ z_Z.vхV;i>򗠚;,K ni2/A hu͓|IWX /2$_)RkC3'~U/ _f%/vNy4/>TY?_/J%K ~_46/ _ $>s͡C.bT b0 11!/6f/WN |1aKH|i vXۅFZh]D֯_|c{mcV>_|B(kGE4'}.34j/^ŋp_ėk?_ |mh&]DOzu >MM^_jvy; βWza%-_I~k&՝,]+9_I--D.-o}'R =yN; 3.ڟU9Q랋1NJ"}2 ȣ-^y>IεJ$6;O;ik?7s{~r; guT.fۙk77p:tȹ9ϼoF.bpcμmì\m,w\{]k#{9(GQg\Jnfi{mx2C3m2UgATYe9ojY7dӛ\&gAe,H-kǖvS9G2JwRwQFQNS'rJX{h!*ʗX^͗^۱qc,cUYǖSR,޻|r|}j^ A4Tm-h-|eҨV._g<@=H=|1W_ |ð]| ~o~ T? vǘ'r/2__ͷ,PR/gVi>C8|7M?/+iS _$5ӟem,Z̻܋̗ vEA*0'4U|YG#V#3q yAf_K 2+G|`+܋̗ j |y/%/6%/>Q,L_LdV/|+Wo /|+W'4B42hDs/j֐E /!%֐/! qᛃr_Io2|ep_%/CY4hw* :.~zzxyGY% k qшՈ溰85EE/%¬$_ 냭'r/U _6"KDͷVֶъ&4D\'4*5Qf_Kk**0|%J·7 ~+[AJ? v7/_, XXʒ'o9h8J|#+Wo*I|y ͢Eͼ ~\: ,/~l:KM?afz* r[޺Udh]زjw|wlۨ|-S=گ~zl_R߷t:KM?a=T(w|wl˩2,,RUfli=7ZKS\ .ͱ6FChܲ42`Tw,:h{KEF>>>:Ƣ3_=:1bે-!5vп؍4.|"""j""=EUdOhN?Q4F:LfWX{Kk/Ckk?_ |mh6܋Vm3 \*vANy4f{O?Q4F:I8 ZtƼ/ܗ!%5O/_I>{j>^'|u·;+Ww1q4zܲ4o\k˫x-j҈_IF\Sߑ+׈>ڟ>We)v U';M;Ms[+}[Tn̍'_i=걳Jk랽qqZx+qX?M| 8qǒJƣ3}bdcj+15ۮr$^Qn\Q?wSN;ߍ#qUFFx2^Rͷ_%9/Qwz_"]{r֣QV%G?7|KT|ƒwnj~vk|`%nВGh*Zbj%|/^Dx\D~I{/ QnO _<|yz;*67j#dVSW-Ym\$nL5+W0le=F5_A!'4U|;ϰ6Fh1f؊.|"!e_<|yz% j -mAhn Odǐr5CfOuRw04Thu͓|>C~+^}__)RkCمO1do{Y/ xWxOhN?A4:LOdǨo?I/>fŧ/vQNy/>C1/_ eH|I 6SK׆f҅O1dnO _'|nw{%?4:Q4[cv%qiw_%!3r wM~Z|.?CۃvVcR;QpDRC Q}C[C;S—/_^o#-mAhn+;DCŷ(,^gQC3WgǍi0|%J·Ǩ+h7|CVo'fȠ=L>ǐZj~fgo3+4T͡qᛃɗw4F)W U|ϲ6Fh]DC|"G Q,`Vi>k?8q\Wtk 2 _KYI|E > +[хO1dlP_×/_/|٠/A -h-Aq_Lfa/&b/&2%_ |ð]뇯_"|$>3͠Ak'{L U4%,BK /3[C4qof_|% t+7}$_YfȢՃfޅO1% QC3;*,KX_菳6FF4ׅ]77/,"K/f%5/W_lE>Ǩ򗈆/_^_"jMV6%/>P2(_,^eVQ_4+W0leQhW_~ T fȠͽQbi,,_,bV/?|AӴ\WA _2|#WIkПem,Z=h]D&aBbUFaSjV/>KM?a.3 2 6F"Ek+ߙE|O,R=WFd3T|gY;Z3+hE{.1 {V־>n޲-Ѵ:ѴHhZl]JYE7Z5RYx,ʱ6FCh22`Tw,ڴhRʢ/:Y4*: `K;|ėAnFc7Z;L<̗0ȥE_PYE4Echkv|茅/,jZː5O/_.|"!u|4|ikuDch 9DC\Y.,jB*4[ _| Vs;r%K·'4B94rhB@_Lb2 11)/6f/WN |1aKH|i vXۅFZh]D_j /A%8QwY% j Ech K: ZtK /ܗ!%5O/_I>Ǩ򗀆|im0/ h uͣq_?gV~_4V—/_\$|Im-% MПcmZwoY _ πππ1_| ztbW[BKk.4jG3'U]JYcV>_|B(kGE4'}.34j/^ŋp_ėk?_ |mh&]DCzu >o͢n/^5qX;Z*y#?k|eﶿ蝕J6rzΏ}M+yrLT>mgjctvw-{?eu3 >?: T2 knf@dcj+15ۮ:]0,_^Gu.3 Bo[߰u#XnT}F$|_+}'Tݷ@w'koGvn_p#fa5 U|oZma7d{B5 U|' 5 2 Vg@6.Te@6.4}PwP,r {;[y7.2 2A<|yw7'4U|oam -hmDso&_'2 *ƅ[;{6_7|+Wo |C [A*gXA#N43D^Ck6ꞅ?'4U|soqig|% t+7}$a͢E0y>א25߮BAݳpWY8khjDs]`&_77/,K /Af%5/W_lE>אA /_<|e5ֶтF4]'2 J1ʔEo|%J Vvc:E"|EV_ϰ6Fh1]%T/!%_%/78|s4C.+iS _$5zϲ6F4.|"Q/a5_Gx½[;{vUX򗰆gm8Zh o\o^YD_"J+j__>؊.|"Q/ /_<|e#D|oem+hmBsK_|"[eQ%DY&ʬh W_ a.|{뇯_"|$>3͠Ak'{.b9YXXŬ,_,  &i7N2|eFė?,Yz̻Bpauax*O NYC F9\0+ߞ||>|uԦ*= T|YуVς"Ek7,)k:KM?aY2߿X* ^E]ۄUrˀ|ê΀|Re@a5[_z̺zJQraKyʱ6FCh2 2`Tw, hRYǢ[i>⫃πππL/WN |1aKH|i fn4vʹ kiRPl  %]Ӓ-E ^ЉZC}hQD@؏i vJ悮qj@Xt+nއ{ڦz=;||/PZ8kk#M5ߎhahw˕1Ay)Hr5WBh*Q®WbZ/J?wU{HS1Ъ.{_>r@ٰ=(/<:dZzƁ>|6̿h o؊$|_?hDl Va, ¿X+V_X | 1NV|ҡ|n' i+E FJR|bU竴b-/VSJ+*/V 1: ־|AON b$|A |_"A>׈/ ł_|!XVh/߀ hV-4wYr@Tka1ÿ_X+Ŀh>|o |2o /Y`+𑦪AFfh7+&_0/ ֊F>| Q/2>t|`IBJѿc+QRh(_#ƿ4´$/& 1: &"c4i1¿_Z/W"B3GRJ:}WroJĿv6@Z˧Emҷkq|cYϳ+JPmWQ|{zg\gsyIi|^]l~xHxH0.a|vs_5BA1av(BNNL?D}$gDzNȊ?`9!>GdTN9#B![$9"._ ̑h yi*w_Ě@\ Bq.Ӑ "nW s@~XdT tT<&k xV;sPn>Vp%L{u7=g9_{ sxOw ocVޟNJl_,t~Λ-|]kq'&[.<Ϸ+_F|1[׭|;u$<_?˲'Nz̢[g7~и]Ѹoj߼^FZOǼLsp9v >mH=ұdsqs3{>wJ ĶAa/5_|Ζ߬pnoߙ?Ya]ΝŚZsghw"K-Ju(hOdI^R{1kJNw_RO&@@o{lPH И遝h4^PDG_Kώ/ήDYRv1%[*A 6|Pʟ?91[-rb6k\=Ҵ ˸ [? [-rAvJvqbvRoIo/ mh#M5蟏ȇlh7QDfb\=Yez:4𵂯|-k_+:V1|5&5l5>T[[ bh-*/:J|sX 9f4d}2 |m_:׆jE5mA0ƆhF(QWlQ[aZX((Wf`,|7녆ZkuiiZX_/V+J׬kWՁYr|a(oVj/蟏ȇlhη*Ok¿/, b%E_+Z𵂭Co+/ f5l5>T/_(2hb-ذ6 Ŷ &/6 . &kS/2 |m_:׆jE5mE0ƆVhF(1ƿQ P.cb_ߋ^hx5~uiZ8_/J׬kWՁYrcC_|QE |aĿ8@-F|Q$' q^=ɉrJV𵀯|-k[}NWf5|`𑦪Ab-F1A>qKqa-\/. k_7Cu)_6t|kkHcCk+4 |w<WP z2ІrnW}@w/l5l`L[ʷcˠQ2-AYfV,߶,5pֻYm(Qvm˒򽙥>ƶA ZmYh7Rʟ|:+6,ɧr]Q*Qf];E?ywZY@#܀?)kt|:ks;J%ڬwg5:{HSo1c7 lξ|!O_:ts t$|! |K%ƮJh-fHr|:&±em&Gj|ѿcۡ˗+c4iߎR6Kb$| ѿ|E+_!4s(QWaW+c-_|%*|WFhBs/[Gj|lX vJYlX+[/[8|V` |_|A[4U G066CsM2j/,ւa_XkJ >|*O_:ts t$|! |_ШV)4C |_|V b]UiZY%Ū;ƶCZڗ/(c4iX_Z/W"B3G1ŢX / "/ 0 Ъ.Q1c-/f3ke ||>[oY_4_|A7 lE>T/4" f50 `_Z1h/cǀ0u2 E7s/l> _H_)W`l4*U ͐:bZTV&1i/ߎhah7)ON#b$|A |_"A>uU{r𕃯|wgU%Ũη0 ]^> |{oZ#-AɟAgs!]IFxϯ턗y;}OxYr?Txj>:[D't(2L^0:n~ L!pgs E3Trǽy6:y6wJlȳy%oD!6φ5x;]_ߜ1W |FcC>_s|B~L3ߙuE<b7c;w<(P.7>wW cgBc&fwx_) j|gFٙ(P.7)B |k3c^T{||*ʋ1|k3[q+n܉NLLLtXHSWf+SӕӕYc<Ҋr ؔل#hR(o6zHSo6cl>45M}:d|7i3PZQeNd2L/_V𵀯|-k[*&F_=&ՃFGj|пcQ e\\)_Yoka>Ҋr,ske#M5.uo84Ys_E_t:Z$|Q |[?ahQ>e[VX((2WXV^>T^B [54b-/V+k%kW5_ ֬G,RU_|QE |aĿX5|hCk64[cT `_XZ|-k_ Z֡ |k_3W i/bhCk4W W`_lX  b,IMA.u |áE_t:Z$|Q |[?ahQ>e/vu2{S}((2X+Ŀ5b^h߮_d|]F:Ł5kCՀ|u`kV\__|Qm_!/u+пc Q+9ߡ_d|7ɉp¿8_2'9VN 𵀯|`P)/ f5l5>T/_(2hT/.qi/. qa\_п|ph."ku|mx _TVclahmfT9™3;fg*WAqn]3LmFZf]aE.,__߆rnW_W >|,XD+||^Ga*h*3ͮbR~Ll~LZ~L+e2J2_"͏qcFfhn@Qʏ5:^vrLFIt|7 | 1Ng_ϧ/:n_:|cWBc%B3G,RCokAZ9zHS0cl;4ڡu}re|>=t:9(QRz#WE+_ G,R :_kB+;^>TEFmF#jޗ-5C6-e2J kesk>|o |l2o /Y`+𑦪AFfhn)V_X Z/, bXa51cǀo:Y"iK|n6/+0*fHrYbUҊ¿X_J+*/V 1: ־|AON b$|A |_"A>e/ ł_|!XVh/߀ hV-4wYcTka1ÿ_X+Ŀh>|o |2o /Y`+𑦪AFfh7+&_0/ ֊F>| Q/2>t|`IBJѿc+QRh(%ƿ4”2ń2IIAvmF;CsI |zt_/F# j+D"| *Q.ҨG땃|[gt+bTۍ h.c/R~̽B~L=_rcȉN~ ,?~,7{k.Oy!;i5KnJB,_'nVy=d,cSgy].c,1;1vqybM!xJNb 3U- {j(P.7~g443UDPvxʧө|P>>?tjwBΩ;iT%ԌX2nȈdTC)GxU/5CώgBLh |E21eR{P6f4m2nɐclPؒQc܈rJiFqQFR>oIo/ mh#M5蟏ȇlh7Q>DfQ)EecDfK 𵀯|`P[h_zHHSocl14 |;!71c-̥(l̘cZ{HSo8w |]Mܗ/*ku|mx _TclahmfTr'|u+,X+K/iG/zxh-}e|]FZV+Ŋ5kCՀ|u`kV )_ت/ (U_ cl>45 EGZ/, bX 𵀯|`P[z5|[4U cl14 Xb+i/6 bm(IMA.u |áE_t:Z$|Q |[?ahQ>ʝ/vu2(l(cb_ߋ^hx5~uiZ8_/J׬kWՁYr'bC_|QE |aĿ8@-F|Q>$' qn̘Z9%E_+Z𵂭C>̿h_3 zHHSտ1CZˠySdť. ŅrIKAnu|áɺE_t:Z$|Q |[?ahQ>Ojujqn]*'O8{xfxE-Sc] `eq ߖj|Q54~ _OmFٵepw]u6(hw|UXyl,JG|G|jG|P.s=+Bckt01cǀo:}B2>t|`IBJ] Z ͐NH"uM(M4;ƶCZї/WhC^zHr5WBh*Q®WbZ/J*|WFhBs/[Gj|lX ({Q}lX+[/[8|V` |_|A[4U G066CsM2j/,ւa_XkJ >|*O_:ts t$|! |_ШV)4C |;_|V bUZVVj/ߎhah  |ztZ_,/# j+D"| *QDh/B ĂHEAmFjˢ_d|_X 3 bZ%E8|V[o |_|A[4U G066CsY0| `-aVĿ0 1c7 l4C >7KOW[ hB3G1ŤPV&1i/ߎhah7)ON#b$|A |_"A>ʝU{r𕃯|*bTۍ h.c/R>L!oy%~N[Fx2dLŕtK8B⾤% >o,Q[f\<>p@TC^I&K*z}/m~9!~&"=i /F},-0r\%,^^V |G=[b~LJfӇs|p{O}oZC.)b>o]#0E1Ӝ`R~'`lL??k="I霋sEu|5\q~*6tdN}\ȟB6Lub1I$ږ<uxJH2^1|fD6^dЯK%0ж8CKgqUvɌ&qmq@p1}.y^;Pa%jO[<~ ³h北gܜ;ߙAb >W%kzn>;O\.ełOzgDMCʣ7/ǂEzqD7y`_rs5nUze d}1{稕i?O̷SSKܜο";훯9MJ{?={Y^O 쁶_88 s9yzk.oB9r]^`>}jJ1"Q<շ%+^4ȣW-W &f:iat\pl=7 w8Yv9[]07WWo xNh7[1Ij7c/9}]c66ܒ]\MwS+'{m=/[M iKv͕?rGmq@Â=c]Idm5;7(!y챿E/݃aqsurǜQ[?P`걿J7sNZ+ݴ=rs[.؟^/_')nOGmq@)FE<\jTo{섴1z=7{s`_Mgssv?jy鱿R^JC2i7:!n{/ ջ[cu N?ikN ia~c / r,7WB-(Wc _x꽐ܜ^`lj9~-({OO:i]==a7-؟  })OpGmq@ٟ۟ ߑO>i=rm5v7,~BE.?{o0cQ[?P/=7Ӈxw:GCYZT< !&*z[7m0 gQ[?Pw`OZFC/IHk ެW3psEJGmq@߳LYNZMY=ºI#JH/z ; {;Z9`z4pIkimA˗f)MwOɼ=NLAă3N#85ܜMY6r- A_# 1%&ԴݟvY)Fq^a{0/:wH_&jc| +ZW5>ˣTS:AGd0jN:y >?My1'sR=:S4}`C?4i7-N;ҪE=VSm5s3$2Gxo3ԗ@ttԺ_|OHk-KSPjtuh\=tT⼛v6+0TT3vM!;=Z6\Nw_O\1?Z'/}ݍnn-9wcNy\]a#ߞ~.3W>Nup韶Zqh栔y,_=(83 Օ\{w_Nw9+,ݩ+;!"rl߀]z=o g/עgww(/бUqPzg+/g1T,Y_EӨ{# w#4+_}yUgFGQkG<+o }j?i_=ɍ?ucݟ=cN=zƍxkTsq;jMe~4 Wp&~fn5v{?3$S?Z}{Gx<ޟ)+sh;y WϹ:a(PnIM2M@ǔ|N9k~Z+j?a_'W)i $'-qiP~rXʓL&汔T#U TSGw~q??q?Wb|ŝ[bjǏf[屆4-~<`{@;3Z/ؠC$O{TSIur0nĝbgCUlkU*Ϡw'q~;1$:q'],ŝ̒_\ǔ'~m[ڽh7j3qo9bLysGÇAz:4Ρc0{tqĎmf9PM6oXyްFl3ǹ؆'='O/6=͙l c)k=б͏}'ET9 squ9&m8ԟߝY:9-C⴨Ðv39vxZ#Sc9j;]lCBlm^ih?բN* &lW`JW ئ5;mslC Ϻm6/gIcu7sh[]m* _◹.)}t>W|p3{vᯅkqo?#NFpu-E\/0G(cHqE+%b]ӯ~?ou/sGQhS<(˄iE<@N[v2{/7WԾ'O؉M)2!vZr5&;musf6q:B--=6r!ԟjK:N5tХsl)F63:9wF{}y[JRWBB.ԟR-PDuhP.(v6+0TRRyB 9v"<׉Hcܵ"zmIotp*.t7+Fǧq#ѵa38LGC(q-ΡkDwkDT'55kD?IYI`\#ڿ],S?=!\#s6ORK,#w^W,;|Z/8ԟj:_q?n(✝8gq!mڹ?έL;7ggs=;&!b9i/ԟjoh?բ1n9bgCUT JqBs 9dc_')s~nۯL(gbC| :OiLC>LO\+{?t19$1|v2g FRq.)y_cz9^lGg84;N5Gtj_'$Zxc/Ԏ~8mC#?64W6^/OƬcl;&ܹj:GM>puU;;SMs :,M=P;9''mqs~sί看ϟI,MLbHlӟR-PLCuø$ U9iUwIc6|\iS|1ǹO.l5Ob = t1A9A QѥBltyOlR2Kiho'Rlt];Rc)h{0bk͝˨s\*W<UXq>wjO58ԿNI bYԎڞ8mXC⟳W~\s֦<=^/6E~6DbjAb"Os9 (QlW`Jݞ^3(FKh~"B _]iоSN5zﯤ'Q_1WcV1.;qoJpq ?[pn? !Y=i{0պXC5gOÝO#pǩT=7tj/$ڽZƞ_<~qڗ i{&?왰ógB{hIÓya{~!1{^_h?բt*j% URy{"B4h;5ޫEuz+?M&K-+=C9'Wru{rpsuO_jKw #sv4^պSqroc\8?0I=mY\DǞsx3Bӵ}JqT-ʳ q[m`>c)SΙ{'\ݶv/>o9{:3^xM{4|7'WNAɽ[9~gO''}r4Woto}rz>Mפd[?3Oٲ_]SM:gԎO|8hV?ɎO z>Iϊahĭi/ՂūjQ Jܚ(v6+0T~U3(ŭ6q_ sh#yu7  f^UB$Y~'T79`b盆5 EqΪgoϣN85n[<#ιᷥ\CsEǣ%Ds`f<hrXra#2.D5^7ecSM:lȉ=P;. l?[nϥ\wx6ċs>KЈs^oh?բ *qNPzTA)9*9>Gџܥa% 떡P6$]rp-7\0N:ݖ?r%6Lcq\=wXr}>5%ؿ ͏x}.Y%WKzQZl֖dkkZܞDŽou9[^we>zTr|ӱ<Ċ;N1S=q[yޚ>q煩rq'xe?/S,Ɲ`kR6N;/;D+S{ۋhJin+;O`l>Gk|>ױ6#lgXDd $Ab   c $A ? ?3"`?"A$:- ~@=vA$:- ΰuaT[ǢDx\s89u6\L m]d.S5qf? [#e.BÖK^/_~lw=>;|_aeD0nf,wcX,9WOOX,(K+'ԫ,+~i!Ra7d//;;Fn 1^Ɲ#i36KB(G-'smnuq-+ycoŨ+7N3v#WBk (?40/a\kuH9/Zn2CO ?qu]Lɢ@33/7x1LxY,ޟ++G#B{ah˥}%4'F 9⩰OFN G _Z&}cn^ -V槳j2Qz,Z-Me+=WGy=|b9QxIxwscvLzdg]z"} kxĈ1_!F7ɖEZ,,K,wpEU-Zn/yrSW^1~sK?oz;+3Oj>5vS]3/^2m^u>w6ڼ5rW<?^ܥ&T[_g][KFeVm(?ncgVE΂"e<.Ex慦mKg'Z7tj/N?omWߥZV0͊>21~M8$^;1\E^͒jq-_^hҡgxÖ1]&w_4bڮ{gddˊ+&\{j3u$Ƴiq+tQmEg\Gm-3f-usZofɗ}kI+:ޤue@}W$fhZ[hG]-swBݷZݰCIvvϾKމğY01މm͌t-Og?bzD/_mĒKޕoHue Yʝ]YfUG܊wyxkWh]+_plrH~,I+?R<+?/C|OYy|?ֺLHv[ijoX̜y}n^?>7ZܽV?%ﶭe,1D죮"spnov뢥3kC}*Io+;yiK/Qo0~f k<?Iparwk<0_<3?:`T2BBHEHEHEUHEHEHEHE%H@%O9 F7AOY@;P AsVsVA!gLTkY W_W__ !`#k4lb6u`kW *4l 4l %Y?P:?-\ #!+|QtKf2|2|2|I2wGQAUoTMeE$ttA%_8BOTOfOxOO??HO?7ܻXuW?YsU42 T*?P?޾B~;$^ ????+P?OO+O=OOOaLqOOOOOOOO__%_7_]?Gh_Ռ249_ﳸ____ oQ%+o}RodovooooooooI_[VK}??? 1ASewяR#ЦUf/ ҟJ/\/n/1CUgyӯ -?Qcuڿ$\yԁ1λ|Z|(g${iPхϠA32O?贁Nk@RFRR#Y^Ç?(:iӿٟ` S odXXLetterO_b PRIV@|rpE fg4o/#/5/G/Y/k/}//////// ??1?C;} DisplayUFDfP h-RTUUUA@ ?I? 3h  ePqYk  KHo:Lw >2QO` Flowchartw`u D(2OVhz~ WThis symbol represents any kind of processing function. Double-click to add sub-page.H D  # =hj0>TdYY9 ͉UA@ ??P6 u` 6u e.A&a8 e&0LgH>v5 LS{5 `7Copyright 1999 Visio Corporation. All s reserved.`_BFS.chm!#22268D" 0d9 l>YUdv "}T <b;2F6& :2ge24 7=H?{1>%??K91{1^;.6=G=E%2O;rA wB}C! ai;(!aOj_(O_k,OqV(@^0A^ 5F@dL?VVG~#A&QR?_\5C _T oo!TS>%[fojoQ P3m5#_d^*+T$]tjt>"e}sIsQ q?B`Cost6,Enter the cp associated withpis process S"`@ zq`vQ:bsDu* wd@ {ofsstep @zq>%$s Resourp:wnumbpdpeop{le quirpktopm܀testask n>!auY=@|v=# %Properti:0Se}tsustomqsce sele}cqshape s{ wBb  f ,f2^*mb?ؿ贁N8i6??54 _ Q);M_>m w =_H'-D !OyaGE_5>FMw#C!MB t"^]fa]@Zk]P+84aTd9pE8= I|* UFDf h-TUU[U@@??I?`d buoqYkQhu23u` Connector ` e1Crw UH ^   -}|5 p`I`V?}`xa833ސ]3σ3u 33?, ?Gxpx^& CThis connector automatically routes between the shapi ts.b?2jZ/0? HD @# =h8T YY9 BT#F oU@? P6 u `u bA@]u  .(#DB uu`h?\hr|uVa@-?bl;'bE-ho'$y( 2rq?@I ?$%? @"U*5L -br  ^vv"(2uI."q28v"uh9Bd&</MS{ #145 `Vis_BFS.chm!#22291`7Copyright 1999 @io Corporation. All $Ud vE \/4 *&1$b24R(^[w[D ZQi a59 93O'2"q?/g;2GHu-!N !OyaGE_L>F=Tw#ߔC"7B uA` uVM J.AJ<-QJ*(.88YFxdnJ>ʷI@>brM J?V,"rZ|-#%))<"<>h>@/<@SNG1L4+Ir"`?Copyright (c) 2001 Microsoft Corporation. All q2s reserved.l`Vis_SFB.chm!#265119S3!tA7 30lJ0JUhh5!*q+#`#5brn"u$b@$ `UD"B($A'3`#N89 _EA$^EA85# OG>:%?RjK_ lP\RV:#{^( bPA,b(EGOO@a`Y?ffx3MIaofBDa(2v#b%d@oA}a5Yi'Z yI@Oi85k {$_gAQA(D<5oicOUwKHa^)y@Q%8_[?n[RVdMle85ki\f"\#B2#FA@CA -@B-J;#mQhe$67+B>T11bR"o2sN G5G1a b`Cost6f3,Enter the c associated withѠis procgess0 B`) %`a#Du3*f3 ̧dE {ofstep0D5{$( Resouri0:̧numbϠipeop{le0quirktoՠmte/task0nZ:LJA7O1G1I3 %Pr?operti0Rb ł01Bce>՟I5z{~!g)dP(:hTA?"aa2(QG1e`Dct,data,dLly,acible,such,sto,onVsk,drives,Basic,Flowchart,inoform2,f~Vagram,joiners,ۥ0!5S $@=H-H? !OyaGE=>FR(#XKB CYRaDSuo@+_,Zo5aZ1PUFDf h-TUUU?@ ?6I?d XboqYk*Qu23u` Connector} ` e1Cw \UH ^   -}|5 p`I`V?}`xa833ސ]3σ3u 33?, ?Gxpx^& _Connector that automatically routes betweene shapit cs, using a right-angled line.b?2jZ/0? HD # =hj8>TA YY9 T#FAoU@? P6 u `u bA@]u  .(#DA@uu `h?\hr|uVa>b@-?bDl;'IbE-ho'y( 2rq?@I ?$v%? @j"*5LA-brB  ^vv"m(2u."q28v&"uh9d&<?/M) #145 `Vis_SFB.chm!#26514I`?Copyright (c) 2001 Microsoft Corporation. All $>U@dv5 !\^4 *L1$bC24R(f[[D ZQiAa59 93O'2"q?7g;2GB81aj e/}@~!gdao7 T KaKa 2(c1B1@Dynamic,connector,autom[@cally,routes,betweenus,Fy%w#O&7B ~'gdo@{+ "st(PUFDfP h-RTUUUA@ ?I? 3h EePqYk  BHo1Cw Q2MQ` Flowchartn_` _ QJ /?|wxpf6lbx{s؅pwp  wwRIndicates a file composed of records thatntainelandset operations.UH TD  # Uh8JT eeYRJ>UEMUA@ ??QNuQ`t?u,AMA JD$9A 9fQjQ   k@QM.B:u\zaaJN[8!@P `rpr{|" Nr\wb#$u8H N# "HoLJ#)1.4+N`Vis_SE.chm!#29529I"`?Copyright (c) 2001 Microsoft Corporation. All n2s reserved.S>#7>K00RQ) #53sV##0O贏Nk?HH&QR ##66H #BCMlJY0JUppY5I q! (w*+ 8"?4$bS"$"'.aA aM  " k(WMdQmO*$ VHSf8*cU/Q&t@ IzW$"d$  !%SuuXf FQX3]OlSQ5Oc [?_}Rc#mm 2$|Srp[AQkDH5I OUKSoowZy S26WM Q%zomcShoBf u%d%7} RW 5XdM!mX) % H&bVt$@bT2rq'C#$b"1w#'0YвXXg !V i!VӓٓW8 j[AUOPas~a}3)ph pH%| | ] og { P`/v,8|Nj Mc[AYK]o~!V_6 p"ϯ+{ R%_37%CUYh r){&+8tgcqA(j5 IfpH&H&SE`nC-!V|т %O}߽߫y\P|%V/#/%DB|5c |anbߏJ#711)1CB` %Properties{c `0}Rb "0@m@0Apm7ɱ+NT!! "Ρ΢*a*a2 (s@t)5)1.3Ct'Cost6c3,Enter the cn associated withis procgess02`@.%=S.#`Du3*c3 {d {ofstep0.5{ (` ResourDc3:{numb~people0quir֒tomte^task0ix0BDtK'2q BBBǢ$Fӵ8KcKgTtAA"!!2 T{.1K0Database,Indics,file,osed,records ntain el set,2s,Miscellaneous,Flowchart!nectors,links,joi0C1e~:0inform2,sy ms,d?iagram!X+c9$@"mb?ؿ贁N8i6??54 _Hu# # !OyaHEhFĒz,E-B .YaDY[o@tk\oiP+d54aUt4:mӶ@b_"@ 6}C-7"A8t4:mӶ@b_"@ 6A>-7U=?@t4:mӶ@b_"@ O>7=C-U7"A?@It4:mӶ@b_"@ T7A-7;U?@|t4:mӶ@b_"@ 7C- 8A @$6IR@6BR@$W7HR@7KR@l8JRH<( H<( H<( H<( H<(  EG9 RE$T9 REa9 REn9 RE{9 R_f(2T ݨS**BV!:/{"DR$/"v+w5?l9e8PD59T AM1AM1PPTPPT1h/l'; A/4; A/A; A/N;AV7V@}@ @g "*^p? N(m*~|;CU3P (v@#@@FD Tey Аahm$T qUI:mӶ@Ib_"@??I?.CUg+`8q W*.%/2?qu` "'"-"u$7)? t2?A/-'/sU!&t/ߡo/?J?a"2c? 1CUgyyAt} G(:L^ewqf7* #5GYk}"@eUAi1surface ation oufiplppve noun7e_[m-//!/pE/W/i///?// ??Yb?,i<-h??oh8OO$O6OHOZOlO~OOOOOjQODnU __-_?_Q_u___^___W_` k77g I$6b ߝrilj|@ 0 M '2md5ُ돩/'//>/? 1rr@ 2DVhzTᠯTOdͯe2[I%hMw@InflecAal GenerB*N`r ϖ5/0_̿T_ϳ/8J\ϒ߶Po]?re?oh?O`gPbtB 3_QS_w,$6H4t666f2a$6H@UwUQpi_go? //0/WOT/x/////??,?>??_Ot?ޏdO*qo?O4Oh?o=/59ss1|%9%*c1НQ> tYN/Qzs: Q#I9sWi{AnAԿ nA 0BToxϱ=[4#5`GYkҎ& >dg>6$Z&j%'A3$ 7 ??ߓCq]6I?\?*72(BTfx4O2O)FV5G;!Cqr9_ \]$.  ?@_E@RKo//&/8/J/\+k/}//////"e'85vr?ߓ:av?*.????O"OFOXOjO|OD6e7='?OOO\:a_ F_X_Zer________obL"g1 >o̫1xe0g a5a1aiT@@̣e?V0etȐ5 K'!(ɶɱs ُ0HTÀt䆿sP^Cۢubb^bCz$Oe4q xKnЁ mdBdI+JTK"h/ԤdBad%Kv8*8ߧ1a鷫i1Gvpn!zsY[7Ԡn@n|6g}d$- -c!c!'lq̙?̯4 UB"[mL| a,gz  ٙ0(9 |L?ѐ[wsXɿܲͽ ;p3P+Coord_cleanup.xml bo,gQ!:o/^opooooo*:m @@6id$de,5}4ruC`uCu4F~2`2@@ t@ ?>3@t 6?HQ!BDP"HF76GHU@mnLO=0P&9 {l~OƏ؏ ӺO5B?/VNb_qQ!_̟]Q!^m?x (lM'̏BaI?D5d"4FX6|(4֓3S'NmXRg]{a___oOojR%mLAq}SDqп@2v#4~wB,6mT365&!/4/yBI'ߎe,?Pbt6"mX^!TON4M˯ݯgAOHqrgQN_P*B*CUgyϥϷRSPU@S4^!mfJ/ooJ1fooo/ނQ)3|3vSw/1^////?1?C?U?֓ 9U8VbrL???7C?dQޏ$OQJO\OnOOOOO_O\s9a _pt9a=_a_s_____*<c KĐT͐ao%o7oIo[k{g 4c pk]o\ncw&$/q)ƹ m1ԩ"(ϣ/s1tŧZl~L<p§B[ 24̑4HZ UE/Z>duaq^/?@ȑZU؟Ȃƀ5]RruZ`LnQ?u7"t #hut WU*&@KzH#[4MsmEE↱!L40[@ɋZ1?DxQQh>$5Zw$G$A/;3T/Processing stagesi44_XP0Ep#,'2*.*qt4"=.Y1pEH E[$P1/鋡EPao D]ds`U /E V؂ br1  @30\MIU1qU1BFWœϞ0'40O贁NkTw&g BGhS 2?_VE+?L+M/塠p[E4phW``Ѹ#`Vis_SFB.chm!#2651Ib`?Copyright (c) 2001 Microsoft Corporation. All bs reser?ved.APEYcKpU1;Zmcm@ [X<+hRT77Q7QE$ztZt¶أT&MeQeL6c,Enter the cost associated withpis p#`"o`Cp?`@q>š`PjZr*c vwduc ofsstep`Y1Dg{*PoD`:wnumbpopeople`quirptopmtestask` Resour dPl//Zq*Ḳ&z&"< biBZEѿw///a+6GF|5J/+XX/";FXg4.@R|ripZZ9Aw"q; Aip\շÂDct,data,dQly,ac"ble,such,wsto,on[sk,drives,Basic,Flowchart,informb,f[agram,tjoiners,u`&/U$@B!5&Mach~-learnp`!%*1E'?9?pj3Ec;TT!3Α r2bvdT?>a)k4߲c";UnPdT&8JhWEnQuS %> perti<Bb AWQ]1W]< m2cQt1Qt?bcU5ʗٟ EPQ.! I|8EІSKMW+ɿFHSO gQO rRv@+QIgT>^aooi[Q/Ug_ySa~0dsQsUQCS[(c8LT?v?-au(ply langu indepent‘5Ο#|vc,|#/U/ |# c/*s/q괱Q@m7緮-)B(!??e]1Ii<8E??zi;X?4/U? ObnE[2bQ&q"ο#oVPhzx7L83GKQ"b=!bnI3aƱ,vG!a=!e"34@@{?@=ӽ?@%b!3 R٨lH |ab AL &W| e NHaEj:3h@GXi4e*`#s4od2GcD?hzvI?dpAB(AOO__&_8\{OOTym1">sp{Ɵ //0/B/T/f,u|/m7!` H%q:vj/gxHs]vµ$3px*<2KDj|DT classifier*1i`r߄ߖ&8Jyn73EPt_=o3EWi{ 9?8ؖoo&8"JY$mTiCapplicable to morhan one our curr sek/r//////*/?&?8?u}S/?|ſ?//O??C@?FOXOjO$EOOOO BVsAj|sAOa)4FXj |U'9K]o0O贁NkEoEsA 1completely German-speccDV_zL/ //./@/?d/v////?/ŏ?*??(1r?U??????@O&O8OJO\O@nOOE5L~OgaɣߥQzeEJO`ەEkDY}͉@@VSq~Bam0r+#,QAdY2!!~/& ޔÜu?Qߪ7u4*qM9u@Y,Qo/ع7пm;A;JdMł&%3,U!T@S<# }<N`W{  M_q vw U!T3ʹ-a9#Z$m,Q _~S_d3"UD%uɏۏ#5&0>vt$u ,ᐬ3P-&IU―2eU!/i3į֯0BTfx%3h3ɿQU#ϳ U-Q,GQM9Q]_ϠϜn6ЧP_c_9%1dq||ŗ%@oo!?՞%~La t@SpN">VoP-DT!w-u``u b"@(""! 41auK`0?F<3u,PG9kRr[o6"u/r A |+!@yCp?36? o6Gn7n7#O5Eo6aG-M~7kO5@^Xt0O Nk^v 3``C\r~2sVI?Uϐ /"*Ioz=2C9 t%y[(_1!L*/S1!kR+Q2%% b-0Va[Y*zў+[?@**egJ5OoCv 0+4hdmo>tq'x^CB-bbb^bCz&z FqSE^!H0m`!5$6oWɔXXCT7gK$t"Cd`!@f5$t|qUAqXX 8zT*AV폟V*ZceoK]ZBp&ڈe8Yhhj2S܄wtڇޖ ΕgT(:LvcK"!!1!ye1!ï6 ?o1!I _2_?]!r%ΟZ7"F^`CՑ2fB$w$xB.r}xo#Bp@rUEdet_realize.xml{Z'{*o<'`!%%ԁ5$Af`!eVah)}V`"n&fg*2cV95` m0g#sp1!5?(i31Ci2$rC 4=u1,9K|1Aau@Ot,tS"q8uޑ0 #p@r tN01!zs z]qui{2rq?@UI!ye&'f<"ARqOOG1!qU1{֩1Q,Z_lZ](,!I&SewyQ_yQ o/0oBoTo+xoookooooAA&6A3Wi{? f/x/Sewя_:?=OaP*=߻ߟVo9K __//!֯}!Dtoozo¿ \.?ORvψeϾs?՟-5curelpro3Q43Qߪ߀ .@R%pm5'9@@ !DQc3Ql? O@?_?> J\cu!3Qy;$6HZl3Qߎ3Q* /qoPS/+=Oa/A/EA/l??(?_L?^?p>6_i@?????j-@Oc)O;OMO_O(OOOOOOO%_7_I_[_m_____9/__o!ooEooio1oUgoooo wxO5MASe/?N%?rwO܏.əj畇_ f_@___ϊ#oco= TVprepunco/;a[@z/rC 0BT1y"4FXj?? //0/T/f/e//O/w?/O?OO>?P?b?t???3E???aĿ%OaKO]OoOOOOׯOO_)5_G_Y_o}WU9߳_n_oo_ o1oCooyo_ pigostj[q[6dW([q+8OuS4ιS,ɖvquWqysr# kԿrzĿznu`uWp`ߠ@Mu``uSSȉ 4prch@)♲pENS@z3 $p5dQdv[q(ZrR8:L1qb]>aߡȨ tiNzY@ ^jz2rq?@I?{ yv&!0/3yD[qa/׹[q8Prf㈽ux@CԂ@C{rv@ DoVzn____ SѺ l 2g$/8S.Vfew qx.Bڿ@ //@/at/}a//////??Q?u8}`3s?Eu0OıWFb3Fb2ҺqA.Ѧ 3 Q3q!qwLFcѦѦ  !!bQ~vӄ&)eqW$ $|qu)3AtaA>rq3AxYcqAǮՆĩǥ.4ѡ! Pete e3q"_sta~ AAO=U`cRU`qASU`ѩѡ|U`3qUU !U"#$%U&'()U*+45U8<=?U@IPRUSorsUtuvwUxyz|U}~Ut40:mӶ@b_"@ lH4K{C-t 7A@\{LRH<( Ek| R\lj<>:{@? =|.PDx|.PU1( UO"D&aU=QJf )h"Ty9 )U-xk %+& U) + - / Q   U ! " Ʌ&Q- H*9(T/#=-PEQ//,GuideTheDocPage-1Gesture FormatVisio 90ConnectorVisio 00Visio 01Visio 02Visio 03Visio 10Visio 11Visio 12Visio 13Visio 20Visio 21Visio 22Visio 23Visio 50Visio 51Visio 52Visio 53Visio 70Visio 80Flow NormalHairlineFlow connector textProcessCostDurationResourcesProcess.3Process.4Process.5Process.6Process.7Process.8Process.9Process.10Process.11Process.12Dynamic connector.35Dynamic connector.36Process.22Process.2Dynamic conn?ectorBorder GraduatedProcess.38Direct datavisKeywordsvisVersionBorderDynamic connector.8!Border Text Transparent LeftProcess.26Process.28Process.29Dynamic connector.14Direct data.68Dynamic connector.69Direct data.42Dynamic connector.43Direct data.44Dynamic connector.45Direct data.46Dynamic connector.15Direct data.47DatabaseDynamic connector.16Direct data.25Dynamic connector.18Dynamic connector.19Dynamic connector.20Dynamic connector.21Dynamic connector.22Dynamic connector.23Dynamic connector.49 3~ E3 ~ E3t~ E3 ~G38~ E3E~E3̸S~ E3`~ E3m~ E3z~ E3,~ E3D~ E3\~ E3t~ E3~ E3~ E3~ E3Թ~ E3~ E3~ E3  E34 E3L# E3d0 E3|=E3DM E3\ZG3w E3|A3 E3 E3E3E3E3 E3$E3<E3TE3lE3 E3E3+E3[:G33YG3xA3|A3A3A3A3A3A3 A3E3,E3DA3LA3TG3t؀A3|܀A3G3E3 E3A3E3-E3 <A3@A3D E3,OA34SA3<WA3D[A3L_A3TcG3tA3|A3(G3A3A3A3E3ˁA3ρE3ށE3A3A3$A3,A34A3<A3DA3L A3T A3\A3dA3lA3tA3|!A3%A3)A3-A31A35A39A3=A3AA3EA3IA3MA3QA3UA3YG3$xA3,|A34G3LG3lG3ςG3G3G3%G3=G3$\G3tA3<x E3TG3tG3A3G3߃G3G3G3 <G3,[G3LzG  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~UnU U UUU;CLt4:mӶ@b_"@ SXC- 7A%t4 ,_ A-$3AJ@LR@ 4RH<( H<( JE,g REDt RY{  g"4FX (h(ZH+@(ͫ4?sªXz [La:YB P  b\#Ӻw&#J!ck>q#~F@N35s| jR$ ;wP.\xN̳ԥ)>B<'d|h1l}===2 T?=*PD<k6%`|dU Cg yVisioInformation"]SummaryInformation($&|DocumentSummaryInformation8^$_1071657581+F }G!G      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwk {|}~Oh+'0@HXdp|mgamonGx 3 EMFl@VISIODrawingLxy ??d(t(ytPPPppp@@@```@@@ׇ@@@@@@@@@@@@@@@@@@@@@@@@HHH```@@@((((((xxxppp(((000```@@@``````@@@``` ```@@@翿翿翿翿翿翿翿翿翿翿翿翿 ```@@@hhh@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ ```@@@@@@@@@@@@@@@@@@ ```@@@ ```@@@￿@@@@@@@@@@@@@@@@@@@@@ ```@@@@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@888 ```@@@ ```@@@ϯPPPPPP ```@@@ppp,,,444,,, 444,,, 444,,, 444,,, 000PPP```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@XXXhhh @@@xxxppp@@@ ppp```@@@߿```@@@```@@@      888```@@@xxx@@@@@@@@@@@@@@@ppphhh```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@888 hhhxxx```@@@`````` ``````@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@sss ``````@@@sss@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@HHH (((PPP```@@@@@@@@@```XXX```@@@HHH@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@xxx```@@@```@@@```@@@```@@@```@@@ ```@@@@@@@@@```@@@```@@@```@@@```@@@```@@@```@@@```@@@@@@@@@@@@@@@@@@@@@@@@```@@@```@@@000@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((```߿```@@@```@@@@@@@@@@@@@@@@@@@@@@@@```@@@@@@@@@@@@@@@@@@@@@@@@```PPP DDDsss+++ppp@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((hhh``````````````````HHHYYY``````````````````@@@@@@@@@@@@@@@@@@@@@@@@```@@@@@@@@@@@@@@@@@@@@@@@@000dddNNN````````````@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@``````׿׿׿׿׿׿׿׿׿׿``````@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@sssPPP```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@vvv[[[׿׿׿׿׿׿׿׿׿׿׿```hhh```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@GGG@@@XXX@@@XXX@@@XXX@@@XXX@@@XXXPPPNOW ```pppC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^0Crsxppp@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@"+\C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^)k#[%&/翿翿翿翿翿翿...翿翿翿翿翿PPP@@@@@@@@@""")dC^C^C^C^C^#2C^=C^=C^=C^=C^=C^=UC^C^C^C^#[(8@@@```qqq#,dC^C^C^C^C^C^=C^=C^=C^=C^=C^=C^C^C^C^C^#[#2PPP``` $EC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^#2 .<>GwwwHHH 0C2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F*;@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@888000@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((``` 46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?UW_@@@``` @@@@@@@@@``` @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@<<<@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@(((ppp yz46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?>@G```xxx000Ͽsss::: 2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F&5```@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@,,,___PPP``` !)UC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^ ,rL,.7```XXX@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@444___GGG!+dC^C^C^C^C^C^!/yC^!/yC^!/yC^!/yC^!/yC^.AC^C^C^C^C^#[&5 ```xxx000GGG!+dC^C^C^C^C^;RC^C^C^C^C^C^C^C^C^C^C^#[&5 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@)))>>>@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@PPP!)UC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^ ,rL,.7///A[C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^4I@@@ego@@@``` @@@xxxXXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXXXXX@@@XXX@@@XXX@@@XXX@@@XXX@@@XXX@@@ ǿǿǿǿǿǿǿǿǿǿǿXXX ߿߿߿߿߿߿߿߿߿߿߿߿@@@@@@@@@@@@@@@000@@@@@@@@@@@@@@@@@@555@@@pqx46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?GHO@@@2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F#2HHH@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@000@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ppp"/rC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^&cL<>GHHHklp&5C^C^C^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^*;C^C^L#2nnn```abh(8C^C^*;C^C^C^C^C^C^C^C^C^C^C^E'6 [[[```"/rC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^&cL<>GHHH.C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^.A@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@000>>>@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@```EGO...@@@@@@@@@@@@@@@@@@@@@444eee@@@@@@@@@@@@@@@@@@׿׿׿׿׿׿׿׿׿׿@@@ppp@@@ppp@@@￿￿￿￿￿￿￿￿￿￿HHHppp@@@ppp@@@ppp@@@ppp@@@ppp@@@TTT@@@ppp@@@ppp@@@ppp@@@ppp@@@ppp...@@@abh46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?GHO@@@'2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F!/y@@@@@@@@@@@@@@@444iii@@@@@@@@@@@@@@@"/rC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^&cL<>G@@@׿׿׿׿:::000666:::׉׿׿׿@@@klp&5C^C^C^*;C^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^C^C^C^L#2TTTppp@@@@@@@@@@@@klp&5C^C^C^C^C^C^C^C^C^C^C^C^C^C^L#2TTT[[[ppp"/rC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^&cL<>G@@@￿___HHH'C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^0Cppp000ppp@@@TTT000ppp@@@TTT000TTT@@@TTT000ppp@@@TTT000ppp@@@pppEGO&'/***@@@abh46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?46?GHO@@@'2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F2F!/y@@@@@@@@@@@@@@@444iii@@@@@@@@@@@@@@@"/rC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^&cL<>G@@@׿׿׿000666666000666׿׿׿@@@klp&5C^C^C^C^2FC^!/yC^!/yC^!/yC^!/yC^!/yC^!/yC^C^C^C^C^L#2TTTXXX@@@@@@@@@@@@@@@klp&5C^C^C^C^C^C^C^C^C^C^C^ C^C^C^C^L#2TTT[[[ppp"/rC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^&cL<>G@@@￿￿￿￿HHHC^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^2Fppp@@@ppp@@@ppp@@@ppp@@@ppp@@@TTT@@@ppp@@@ppp@@@ppp@@@ppp@@@pppEGO&'/***@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@```xxxHHH@@@```NNNOOO ```TTTOOO׿```xxxHHH@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@888 @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@444:::@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ PPP`````` @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@>>>@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@888(((߿ZZZ  Microsoft PagesMastersPage-1ProcessDynamic connector Direct dataDynamic connector.8 Database8_VPID_ALTERNATENAMES_VPID_PREVIEWS_PID_LINKBASE A  KYYIJ36b3q~O2~8pVz@u]sf?0xwC^<{gޟ&>}kILw21UUsTs`8 }x}י&+W dua{fm NLk;ė<&b\1xē4%e^πYk/fI֞$C~x=N}}߳3G gp?v~~bޯu9_d, k3zps= =u '/1~'agm Od2kqdO+ޏ7??xKϙ}% UWrl% f/mfm]d"kbu#qN{ZN2^b\{7Ue kો U(+W [f{kXKֽd udžvŚbًXz)^b٫A4Y>&ݱC}3t/X2cɎ1x䛡]@9j]_셝~/v^}"Ӵ_G8BA2[/1E C/1U/ ]j|W[os_|A|37_'2MX]dm#ssP];/E4MD/UϮnો]+W_ 2lWK>i/_ڵd%^2 %J/Q ^D/QK^EI/>Yoe/:_@FL;v# H 9_d, k3 |p6~6F^6_l #=Bj3H~d/"H%~d"~Q}w5Oc+1mfo/c[ Di0kwmdn4HBDKADWR(2e`"t+W_ 2lWK>i/YYa/V]X^X+b_UB0}־OdIfuR (ȱi'cXW]@YiIL,g2/qΒT2RJf3Fx|?0Gftv8gya~|1~f\voEoMv!|{Ɗll3$~OI}%B cgo?̺û|6"rC/ 8C2 M:_S:ʇςخ01]/nm˿,^?p+^?pmDu1.r:H|f|F]qR~tL^clU;a ?'4L7 +_+٘q|hO.Ź"\QI| ?&Kmfr|RKqbg)7GECCX ˋO&~Sׯ_{cS,ޯ3.2M2ʙ1cCV9~=%Y}ou c)M+:ю_;No5ܷU|{V'2|̟Yd"E֡pߘs"b c)*b'2|o{ k^6 :_r;8K U:kNZE_adoVon${wn${9f,'6$,ۣ{Swx$voŷV|cf mS{o {IbSa2ƐUN4A|IQ ^D/QK^EI;_;v|7-1|O{ߓöUL~a,"cYd.2h]D+K4{MD/K4{-KB{ k^6h~|/sq|񽀭]mP}wn2&dn0%7/_b藘bث_bIM"#,2 c^SCN~/vzM=ۊo;OubW;{o Wv_ KXɜf7/)V_bًX%~_b٫X_|k׎ű~Q=I|O{cضJ>i/̟Yd"ű/qEG/qU/q 5|{ " EkW@q|vɷAMڻɸoWx~g/0v0x*^x~a~kH"+x~^SCND@$/#cM i~\쩼____SyޯoJׅ260 k d~S|oLJ~oI|g{|,=7?r"~'2|%Kjo/,|Yc-|k".Ym;g{0efثA4g)֞"Y<7W%*>e/bcߎ}A,{%|Ә=]odž5݆_5j|+v'2|O2md$ctSvso셝^Kvz^}Y B|7_!,*_|S+Wo*%4oa2Uv~ (K {C/1K {#KBKė/_c E'' _,|FKj"ڻȸEdVE%Z_ًX{UK4{-KB0kOqd6/QA{ED/QK{%|Ә=]oG%J_UW6|wEIe6ֶF2Gwf;fc/l~>fclR( 7h|ز |l~QMW_9خ|"Ӵ_[B2W %R/ ^D/K$^EJ/%KH~F+—/ _>ђZw.2"k]D(K{AD/K{1S=E)a/:_+{a_~W\7 oU/V~W_5eŪ/ocmmd!sՠ_t>q_ie/hNY~Qo4 eVY}3|S-"~Wo*Od 3v [@*P<wen(gV}őe\d'kf*#_Η5zox1c?Uk,{~6y}vۏo?v3 y|F2n$FܿU:d ]Enc8d ;"|'FZkx'qk>KeQ-5Uo ]h׵kwvͧ(M U v}<˵ۭ׆]_]moVhvmЧԮfլ-~Zv;%v/,Aw;ݵ~n՞ ];ꎿ_7YX%%M^^t)KMeH5ŵ3TKχΚ(jծ,^sלz2xx>dq-s_}xu,9˥]Fg.v+5~YdovLIòO3N&yI?e^IyO'{2b۽ɇLJ,m5=I|O{caf?TYd"kVvjr5cq/3M$OM;841'"|f?==?2"|z_) W&W%W2nd\b\&e֙d1Z.,e}t>QwT2nd:_;f|v7[0׈_#V:}"7X; dM!scGs R؋Jƍ_A {2Q׉_'z |R} 'w|O`k|"W*VQEV RR}R}Me/R+72j|MeR}"w_>|}cKHk:|Ӄ_WuJF:dJ2VF_i_%H_4%J%M_b Yydק: g%Fd$n7k׌o7p*Bk׈z|R8X; dM!sӠ_t>WBK:NWR(׉1l=e~Q׆ _=ؚ%4W*j\n/~P "~ɠ_2 *C ~ׇ_2 ETBN|=^)|2kWl4f%ܷ(Ȥ_2ߢL*SL~a.2\dY9צ).Eb$_o75kƷK/.~׈_#V~q) Y[NF9YSȜ2ϕ/YEEd/YU/ c:uw [oY_|{k׆of'2MU"2e;WV7{_~qWn_ _>|%F+&Ȟޟ} c:yL,whvf*L}7Hlaf]KZfhl#߫p_wa5YjW[UX{sdhC|~&K;j6e=ݙʸL4;SyvSsda}AX'ljωωcKo!kY[KF-Y :\o[zŻƵL3w.+ ES-޳+SU0_%J| |MLs||A|A|A|fDm%22:_Wp{ccp{1s*+4g*|+Wo2J'2|k2ZKf}Pߙp߾t~Ig/t%~Iݗ.KB0ߎώώg E+VqssXŒ/[Z֒QK2_{O1v0xwWN_ .vE2;|g؋4%~I_+W`~%J|*t/M%ܷX}d")͠_tTs˩E*Jn^J +W_wO_|+Wo2J'2M m dޓj/)~IQ"~I_R*E~a~ |1Αy&Š_tb'>'>'>?U, ̯em-d- 3h{ΡB8 _􋃽r Di0]dtu(&K2L$WLWJ|*U`+35%%Y_W,KB0k]d6%'m']dS-Ig~Q+W/)EN3~>وn#]S&/|}ǖ:רu*^!_|`| u_ڕd$k>q6$_%H_4%J%M_b Yydק: g+yd$n7k׌o7>q6Dr* F|Ƿ)S_?Ș@2g: EML%H_t%JEw _'N|a1-Ko6|m,Di0UdTUCt~KBdKAdWRd( >|}ǖa/:_rz Nר{XUd=HfO LsߢL"~ɤ_2 _^eJ/wE<2K3t>^_\kSf~3f|vck3!aR|5[oKB0d5̙.~y~b/,%~b~Q׉_'cz |˲tۃ _6|{5K>i/̯bmUdՐ,ˠ_rB 7_􋛽rKV|};-=cxy)Ms 9,F5SnzXGֺ᣹?ڽj#cYlJ3O71^/Kjb˩U/ b|9 |E7_2|e&c|"Ӵ_2ZK=~Ia/R%~IaR~IQgR E+VqssXŒ/[Z֒QK2>q6$_ `/~q/1.vE2s;^$/K2^I2_+UW'ΆKB ⛏6Yd~a>#cYlJ6O۰Oɤ"Xt:)l/ b|9%yyI&ߌΛk-Ю;2v+wߎڼh΋<i,ծ/r:$UKY]veL}7ۅ봜/<;)ߪnΝ$?ݶn\;grs<㛼,~Dm}ĥ=o":o}Λ|74]fM*]mү&o/bM7^b Yyd2ΛȾ>qWic%یV>w@5ۍ퀁O }+5kķJǠOd0k'1)dt |⼉ HY6ckiA {2Q׉_'z |R} 'w|O`k|"W*VQEV R7!T"u%یMeR}"w_>|}cKHk:|Ӄ_WuJF:dJ2VF8/i ^/iK^I/wE<2K ESYs% ^I o׌nl |l[T|5[oSB0k'1)dt'ΛK:{N/K:{.N|:c[|mۃYL~a~kȨ"e ~`/2 %~`2~P|};-!à_tN+׃lQ WvzF8/EE&Idn-]^eJ/wE<2K3t>^_\kSf~3f|vck3!aR|5[oKB0d5̙.~y~b/,%~b~Q׉_'cz |˲tۃ _6|{5K>i/̯bmUdՐ,ˠ_rB 7_􋛽rKV|};-='GH,5;/ G]ØŨ~;T>/Q^jv^bZ,0&0Fsз}TQfUW6]dQf'Y 2~A/FhQF,H3oFNpy cS#e_h|VetݑMivݑMYo0N2^3뙑 3#Λ<3~qsda7} >?9N|N|N|~X| _Z2jZHf'ΆȾoݑ7'/\zfdk'2|G.2:BfkL+fo0N2^y6|e WW[O }L3_ [YJF+Y-d6e9u>qw8xqUƠϩ +W_WTMW _*%4e~kh k-C}lۗ/E:N/όܗ.KB0ߎώώg E+VqssXŒ/[Z֒QK2>q6$_$ő;+/N~a~k"N~ iEF/i+SU0_%J| |lHX) jӤ~I3b>#cYlJ3O71^/K#_NeR~Q+W/i'U/ +7[6@Z2I5](K {B/)K {"KB0?_ |q6$____||R$+ vEfSA|߆|N&8xqgFMEW_163:ozIbw}k xDG9or*t>B>o"Xnre /9j˭nyW?ծYR;q\kWjgH2#Eô/ ]dv6F;K2Co[;+R]d3'߁3摰$o,X#ό}e_b1Z:!~|i#{BA^]d d9[2<6"<Éf0"&#&yiɤ?0N3<6<v4"50:oqeqoI_cU|6[ŧ'X]Yl"2**NM˸g+1${K&yKN3N&0l%ES/ŷ^ldXlt$$ՑYebXSK$n7k׌o7>qvCtk׈z|+>i @91ף f"%1Q^R^ z|Gu׉(߲p_ |wؚ%40UdTUC沔>qvCMH5}6Heb|6JL3y|};-!uQTBN|=^)|똿+XI:2 |FX)K{F/iK{&KB0Z.,M3Ogu\(d$n7k׌o7>qvCr* F|Ƿ)S_?Ș@2g: EAL%H_t%JEw _'N|a1-Ko6|m,Di0UdTUCt~KBdKAdWRd( >|}ǖa/:_rz Nר{XUd=HfOLsߢL"~ɤ_2KeWRd* ]u"+̡6Ogu.E{%|ߌ_3 |FX_#F|[ť//gm9dM!sˠ_t>qĴ_؋,%~ɢ_ث,_|u׉߲,](k׆ l͒Od X[EFY5d.2wܭ/nM7fR_>|}cKp NJ34gxxkܯv7g8>2+cF:_/F\ØŨ~;37w=*|UdUa1ᾕ#|w mnw'cf67M0:iDyM#΃lєxάy#Zyy#΃aw/f,gq3 X'ljωωcKo!kY[KF-Y  2qmQZXXcOݐ}%H_t%}ވ}R+ Αy&ݠ_tb'>'>'>?U, ̯em-d- 3hg7i߹i=Ռ#wWN_ .vE2;|g؋4%~I_+W`~%J|*ٍ~IS|A|A|A|զIf}G>vٔf/:8b{9H_R#^NeR~Q+W/i'U/ +7[6@Z2I5](K {B/)K {"KB0?_ |qv#____||R$+ vEfSA|߆|N&f,g,q:iވIR(+|FAZBA`ծArDG9Ru[-[,RK,s,uXx+sݲ3 /km~Δi֮Lx;t$^;W;/鎜˴ o^]amvdǾݿ[Zμokۢkr߮ɓ_J>ڹno,ݵH۞О +YcY$6Fl Ψt3(<{Uͤw?9oxMVl"BfrΠOK~8釽n~ I ]_c}_=bVFo8mC9vOWdk2k׊[ )Wg3gm YA[;=u6#=*7>WS=d=w_/^|2Qz_/\+^E{*be, xr2.268;ó11+ڨ.W 5)Zk׊[f6HFdy澊k*g12DUd5>{'2|gw|gٳtruw[kPa~=kɨ'k b_/3-Y%Q }o'2|~;Y$IbP_׫;G?1$+Z|;߄ _d*zo׀Z?{L3Ud'*2+C}]:3ܗsmTI,\ƒ*Ga|:u;W_ V|Z`k|"Ӵ_dm Yɬ1藀_ `/K~ /* K@_׋,6{_truw[kPa~=kɨ'k \]*K.{K/K.{;|'kd8Yk/:_9<%~ɣ_x^ת&|M$W<~c/5k6ʓ%O_?OUdVW|"~ɧ_|*_a|:u;W{Tׂ_+V|-|"Ӵ_dm Yɬ7](K{Q@/K{U K=|z{_tQruw[kPa~=kɨ'k B]*K!{QH/K!{U(KB0Z'Nd׫;G}2~FXUc?Q|1n%^݅)#n/m~8҇iH7( 0l-w3>>3k|߰Vv#h.\zdH3ߎ3ߍ<Lja;OEʾ#|fm$H|]?O[칅 3v19!r 2)$qe^28c}Dȧ 3v1BAC䉂AC7p|9 | }|78|M|"̷XMd&syPp|sߎgb|?rG>{?fwogg.f||r<l_|78|M|"Ӵ_Dj2 %G/9 ^/9K^H/̷{9)<899yx"7[w v KȺ~+􋟽/~O+/~~a ֞ Y/ߠ_t>O}􋏽||?t|M6]_| o6f㛋oOB0/kϠ_t ^d/K6^eKW_lll](>oq Od 7vZM졾׽~*􋗽/^Kx+Od mll%uA|E >9|||HVkd|[Ȝm{1~VOC/~a ֞ Y/gЧ?U#,/Z>k 1;2o ]$YtS3Y:nT[Ңq Io O΃\]S_{jBݮ;o9}k õ#OΏ<}߶|63DV܄o6|\|Hs5xf#r5)xxG:#Dkg<~|CxH:!>K3 P=1{Qy[{1 r+r {`{3?kbKsb zG/=w5T:N<N]lj>|2wLnr>Jk^[Z[}еBRj[VhHw0q]"nIqni;s\m|YB5RXV钆PdkyExwx~灷vdχ]RP;Z.:qW@ Ҿ7;KMc/k#}}4ϝjҾWꞪ}hJm3 ca[,x]m}^¯;KO1slC)u㊏O論[<o8)#ӓ3~݇45\vxĘt)ItMw~\D7m0w5@|_p+~>:.Iڟ>?5{% 7isW\m:SOt;'*:icU{?[*>?/~ko:*2yOKo!sw~r=ş ^uwuC#}z}KuwVX :'X_ ns=ښZ_>s}{=G|;[ѳ'_3;kȨ#vZ ~3^VFo˾dm,d۹ |tnS[d#q[0.S_k7ؚF??\;boDN xr Ŀ]r:< #"d__+V|Z5I>i kdɚOfwO m~_S`dTdL3Y|gwYl졾Cw.|]!|k_z2ZCfO } }/"`d,WA4d 'Y~2}C}:_9vD`dz%_o'5k·VW75[g>i滊Y;d]EfK?`y_S`dW9R(ׁal]](Zk׊[ kdɚOf]AtP{_K~ W_ _/^|gu(Ӆ _#:$_o Y[OF=Ykl0A\~e/r\%~er}"Ӵ_d 'Y~2Kr EU#ǞG/y +Z|;߄ _d KB5[^I/d'*2+ E?`/E>O/U/ :uw[.?_+V|Z=*Di0? AYo/~)P~)_ @s{z6{A| t׍(נ[z֓QO2 |gP_ ًB~)_ ٫B_ N:p'Р_t^9ra0| 3Q|{ߏl5s{7; %saKþ_>ov.־HƋd.|7\{ߙsa'|~O 7]J|WeᆾMpWL}7o:n`|%̷g{?"Gd]c|?P3:ƍJ s >ke^28c\?`{tA FF ߉AC7p|9 | }|78|M|"̷XMd&syPٷ7ܷ#ƍJߎ|*'2|2߆φφ]2O+<899yxo/a2u |gdߦO\/᱊$l-_%d,!2g __셟~/~^~+ O'z̧|^0nd/>JSMbt|M7 8~) f㛍o.%>_| K^'>~5K6{Md/K6{-/_"|E9 |˳u q3ئK>i/Mdl"k5˳ W_셗~/^^y}"Ӵ_ogg.{ E+RyqssXEo_#kdBlٷcSøQHQ` 2NO{}F?V̞cf/+fQd`R5[-_b}y vOY1`v9lNycr̰{VUݦicy7ڹW}iY?iygL-52KEq^qy|^fN[簺Θˈc>&/%x՟1j3'ﶽhj"bjۿ5DYBgm~{'2|gw|gٳtruw[kPa~=kɨ'k >q.F}澅>7q;cEB{L3N:p'7ת*ΑccXw| oׄ N2[ |\ | WA4]Ox"?ץ6fܜ/cX^H;_|urtk׊_+lMOd $k>u9K@_E~ /%^~ ( ^|zfס;BN.|]`| 5̯gm=d!'ńKBK.KW>i/wI,?%*Αcϣ_<| oׄ N2[ |\LX)K{_yU/y Ox"2Ϡ_t>qƴ_ً|%~ɧ_٫|_|uׁ0._]~Q_ V|Z`{TL~a~A2d'.ߠ_ tR/ER@/U/׋=l~:|G׍QlA'5d6Ĺ~)TB~)_ BPB~aN2d,)4Ww0e2~z%_ߣc2[ |l//rΊO—s1↽>ȕϊΊȋveyg݅#>C/㛄oIئ3?}Gz$? >8Ljx }^Ƴn/_*Tl1wFgm 9]9%0%WEUA2<0 `蔔)2bikjO&Zw[dm ~>AiÒn!e=ߵ|H\.ܫ"{N;|s|#y֥k uQ,CQ~AWwnJ~ﳮo^|K_2`{2T/K[4 QuKQBh{Q\ʣ̷sUهҵՃ _$"w͞|d|6|6+Yėo6n452|#+q+=F_!Ft=F_HSP Ъf=_@$E>HW" t_2 ńK _6 YKj#MEA*ԭFC4I_$|k/. ₿/.K/*>?`$/ |_2D|/(_%(V!4]2S8Nℿ8WN8U G/|;fH*lуO>=l8_ėo6n452|/*%}I$_W"IT/(߀ hV4keEA׷}H///3 - Gq1aP/_67 |y8T Wn4QvbWvೃ>?ೃM/×o 1 /ƀ-CG%[hB3ޞb%A$// K*4# H 2"AG>=\~_ (?uCc>fC3[b|[m|wb>Hg][m"o@h4@56>XKksuą{h?QkC3>X=CXKkD,d,/H~-Ir-w,"d74/g6\_F:L8p/$kgC"gg;o7hZdlhhx1SoBN!oCWaIqاqWkkDG㫷5"hUU|_%E|7姣thLh.JhGbT2ߵI=HHk}6>T; #;#; ~|$|*>N#V+[on4 d(B7ݡw}؃TӁr_"Q #9ўZע4t%AGjƹZD|*^@r𕃯|/@Z)||JD_oցO6>T |*B#Z!hNJl( %$NzF_} zՂ|[ "_TU|_%E|/(?uCc:@sS_\qq_\% _ Z` _ F5#Պ֩[[h:>$H$_/I諤6>T7Fh%BsHHZTN?7 q_p_ P|+ Ьx0q7||r"qOEThB+InQxⁿxW_-jW z5-U_5W JUHS_P~:Nth́"x%U/^ⅿx/^W/^eSk_ Zw l2"Uw:Mk_VķN*/@h@k4Q}Ou0B+C|2"kQw:___p_ (_ Uo4e(`/<`O) ; K&1}J1Ek+BZ.wWoiiK>5 ܎Wwī__O&-|Ł/ip-|w}иZ[r11J1پ$Ch *h Vn4eHr1b>P y~oOd%/|e|*_|-CGJ|(_%(V!4(BWQ- yVGJ|'P>| h6{eKU٠|zpd_ (yȃlhfQ<ĭ}NA 1އ+wi*ա|6@ZuЬq H"UHJA 1W" t_2 SK _6 YKj#MEA*ԭFC4I_$|/. ₿/.K/*>?`$/ |_2D|/(_%(V!4]2S8Nℿ8WN8U G/|;fH*lуO>=l8_ėo6n452|O/*%}I$_W"IT/(߀ hV4keE(﷎$ŁT|2td("_*|_6f/! A*ԭFC!/>P;_;.|v|veQ7|eHS_PuKQBhQ(Mhg/oӵKmK(/GN mMP7BNDž؆iZpN?ztZe 1gu7qDcD8ҶP%]x˹kt\y+bcIbc2L}ek||t|Hlݾk]Xjo-9k]ێ֒·\zGqK5džR(eұ]p[cS!_>gyL.I3TDNEo(rj웣^PslNM&>?׋ǖyъ nuY]WחQud_i]OukO&os˯kw?WhVGф.&ZUƿW ;=r~.o˯KT?cUW¯KY@Ǖ躴_^z=kG(YDё.9Ww_$墺Os_[:3;uWV" ^!wy幮L9C[sj_[?^0 ^W_o!hJJ`ì++5x.^M}ŮV{+'G78zFyD--ߣ_{{glqy2/8Kg+zkS 78G8Xy{5ɌFh!*^++]kS}wŚo^]&]~Zf7n]x݈In[%l]W(͎Td_:$?7Jz8__I:.g/M=>vj[e{N oPʷse~m*mWz-HE>ɩ`l{Ӥ_ ~ˍie|:#$ߣ-2mANwqxg7]~>!!`~fzOeZ$8Y.Ju=^G&mJbj$/g{ϩ4qYЋȭqP]Er,0뢱8q~~6}ѷ A:Y?WkV,W bm` ǻkƏ;[ǟ'*l{Or`O~Y_(7mu1K ǻk=Əu?˵?{{ @qzh{٫Ը__j,3\Gٶ /Ҿp; Zܽ`V t/x 3o6i|q3lFJ#.X\)^3?g" ǻk,cџPjܚ{s)_3{}G=C-wM>K`rQ~1 <̶CEewuhZ|1 v_`&DgXPO/}ޙp]HL4h_8]/.arevDy9hM+o;˵s1߆bٶS/7iOy\/{ӵbxo ?PO M VioRq_`9Ҿp "’F?9mgCm=8Gp&;"gɥ_ٶz?CGwbP1S{{__ )55bJeeW5>ߑwRCsvW?yƟ\R[n6mӭ*4dB>x2):_a}ú{\y_˶ř} ǻc\*6ͷUyL}Qh7#܇yq)Ӵ7{͏ǫt\_ ,ax#?{=m/%}xwy$GqlxrmIqڔϘZkiyzVq|='Z="Pjz&6 JqP3m3G4 ,c5.CL-gKhIdܐz;rMڒbb)[ȎO!9._'|o2NzQ{M)]s1텋 V.a5.~9idt%+"!='~\fF ~Y7(}%mvG{Jw=H2+R }5WkjCݕ y[h?>oK,1xˎ bC&/}Z)O }N9w:N+4q{g=xУ3ƃJԡlNL?Cxgҫ<ܝI^.Jtԑ:3"EGJ1}MBmGͳ /|G_׉-au(tơ^xi(g āaHŒgGpqw?Mj aNGk+]9_CЀZwp16inE8A3= z_8{W? M}=hp+>(6zo"A7O~ÎQ{^Ǎt}wDc.s#;Fh!xw܇t~kׄ]]P' cg?gЙgA>Kq3geIlEc ]ϟCNiI.'}ʇLJbC]?y(ω# /?-"^y{Ϲ)wcܽN?[+SsM=J-WSga\S:{XJ;R 9C)ǐO-I8":@rLJ1*ē96p ,C9Ly('KXG99`/W?dzwQNz8Q'[=󹩷}ܝsN0εax{pwp:8~=\IzǠw zǠU$۠wzFԹ&(vOƆ$}&qk0.u35Eߕsy&hMѶQ9(b߱u 8:-T;𵀯|Oߞoo i_wl{E|UB-FVAs C8 2ߝax); +Ci*9Qބ&hC{j _ 3; i<EWoW|jɱ||c[[m#M%ʧn*4R|M>S C__/q8U_TՃo/o/$÷(N/**W jUl">T31Zs(=__*ň0_#ň2񑦢|-k)7o i_wl{E|UB-FVAs _L*ń0_Lń2Ť_Pބ&hHZTNQH/8WZ_*2|Vīxփo=VfY/x򩨛 Th99^_$|&F0 _ bDC^߇`k[HE | |[4g h̀\h.bd, 1/3 1"1𵀯|-;fUw:Mk_mo "_+ZE\/7"2]bXX,"7 &h99ҞoE_+jR+W .VXU}[VU/VWUи Z!hN[ S7Gkuyu iYi}kvRf DZSf9||N B7Wo.rh-7/R\Yz&Q[ìbJʐa3J|M{GuYvCkwng[ʑ 2:2ƀtHsF&O"A*g2${Eoe; \oe󠥍4v= -f2_y;Sg.GZT416> _|]6 _9o K_:7lD|ķP eZBs{L8F )GZTA_1m|> 3/ lǙ| >?tೃ>?"\| QA}Bk!4seJHk xU|i*}ͨ fh} ͺ|> >t/& sD|*2Q~o2.)/ߤ_L /|M˘|_L*hNh>o _QFbWF o2|F i">T/C2hAk 4 2'8_/q8ĩg|Q`;'/ >?tೃ>?"\| P>u᫋̕K2_q," -dcW"U/(ߌh֧Ьm7VgPq1_p.o LiLh{  1/ \7 "1Ah@k'47Y_ =E_+_T _|7lvGQ7|c7|c7lD|/(_e(h>p_b/1Ĩg|Q`;#/ >?tೃ>?"\| P>u᫋̕G-.GZTl^lK AfmF3>f]tt /w'ZSiװ֖S ~.=bvv535pt߹MF/qk ϯU6Cڠ55Bk^m/;H_2z;_)܅VV^CkЌg4߭O1_D:lc/5[,fxo6b"߻ɔCOZe)K/|Tg_T/Dk9-,5[/K4G KqYY\fGP'']: 륜ּX&i[Syi/}N9w:jk2I l ǻc\דy4oƨEhS\ݝY\{bD9L۴[rwr?3Wn:R_f%;7`+0͗eȏIgGuvG&lG:%*rEGV!}H7Fϧ}|^ijP;HY"D,I;'oDd ':K%b   c $A ? ?3"`?"*o ׼ \S[_=o7@=n ׼ \S[_=8c#.nx |g9ɱ$d&@A A4 RQAB)Z0"TP4(F44xHE )Ebwvd|A+;;3I(+D:BQ>*(}=Ţ٢"1RROv(?E*=mUP.&rYHv9aiLUC>@m5"}]F* Ay:ήK(oZ?wuU8g򺵻ޗwysB_xC Z!}c+eC!!É@Lm?6z=^c1Qi?ǃDtq`Du)]#B"z : ~šQ"TzbL&?7ru%Q|\|tEig"4ᾯopDk&V\(bapEߥ4 &Ѧ:_s]iw~]#_GlYuR4`.RnPz?^5⾃usS%oT*c e2Xr1>2npȧ)OREi)^XxٷtpGLgא5sy??r%/o9ϷyeqoU.x"q]kq/3OKo5Nd'OOYX8E奃)J8wӉ7['4IQN =h~؞N;}jDžNQzH]M"'>X@t:n-RXTȣ,սw - k0bhM[5C3pGp q^8VdrX8XT\H'^Q?~lM6y mqlzc`aKOMq}$b >^NcKh˱cK ⎰w/KVaiw(mG3Ԋ8Ʊ90&%0>wB==L=ܛ5kWGnf]ha:~rtpn&e-kyOyNJH.ދvE+JꠇS?/]m[YC֡m~zY.UMmS1kޛyO]<ۃNAWj}f2]}9GLL%,+C%Aͦ4J1j> Y33F25C)kmo|`{xאoHF8_)V·-!dR4J*͘ =<,~߾i"?hB"GcQ щq!^!:᫑[MR·hqMvrR 7YNjlf|/`nhuC1>_˜5+[r5 _|u}|c]|K%W_ |5mOh %?shu5+6:Zd!^!ʸ:Ze7 M3/; Q%l1ٍ$Gg|iI$2s'1h3ȭq/nMqK ME#- EwLktb</1 _j·FƀG/ P 7YjǜQV1}<K͑!^!j W_|;ag76G/|kુT'4MF96_ru+/"_r\j\sw c߰No:>vV'7̝df9πb_^ŋx/6 MS!?hBbI}N/>:᫑[M2 h'/>j<7G|I -@b4}KKZ?_Nૃ}d7ηLo |5WؖiߟP4Gv XD{܌M6:ͷ-ٔ6x>h=1 >L{:݌ߥ?Ol"tt3B_Y0݈)_I{DlcyE_OS27=J*ܔvU$lDDzV/Қ֨ hTUf9{W3k>$lDDtiU>i̭G|ט/_g;m~$lDDIiV __W _l||^sW_9|6 M3Uo`46 |6_}9;"I؈wrUnM _+Z >'l6||]ˇ/|Vf|3ȯbnUh@sbnc9|rVP'a#"WP>iw|+|VBpNcr__:6l4||W2J*,7{ݣx/C< hԣ=JpJ$lDD 7K_ _)|oحW_9| ҭ؀*4 EW-/"_lj >|y6lHu/| [Oh ṶB hNnK/. qQ _\Zo Sh~2__:6l4||W2J*,7{ݩ8q/NIqJ ̭Gh4UZ8_J×/WD~)|W[bG8hW_9| oC/ !s7Uh.v./vja_S+_$_|l||%z%W_l5|B_ȯbnUh@scpsEQ%FEly+Ϡcβ<; $GBK^71 Q*} w+Y2j/H3C~ړѧqEڳ{agu2ءXsu\q{^uyR{=Q{E~;W}>GlSC ~A\ɓu,6jG|5zDwD:} 1;ھHۿt63g[{@:{@V$o@Ghzz5['b571ٓÐb`Om &ʈKI)t2kɃ'$XJ [dʷ{a[I, b77q3Z7L`y}T_Dj8_T3?ML%,ۧj"RwC8jrujxթf= W!n!o+ܕjrWdNFc2Z3ќGh;ns\:чz>i"?hB"GcQ щq!n!~kH&)|K[ j4k DoӜoZ7yrR+g4F~s (@}DE-\}[r5 _|u}|c]|K%W_ |5mOh %?shu5ZlsEvbx*Oh} Qw/an7OW'>:>އN7Oo&;hDsq/nMqK ME#- EwLktb</1 _j·FƀOkG_<y̓oZy41+&hUfD\zqNɡV9 _|u}|cst"jુo lK5|B_hGk(cs %W/K-r\%ɥV57|;1 [L$v}oluy|3ɟhLFk& DFx%K-/^jmB~*sSHE˅^_| uLW#e-o|Ѭ1ao|V>$QV1}|>SS ?_Z> _|u I𭁯j[2 43?Xc>ѳð+Ŭa[RM6wh{RzZn{R'Y3ÔKK`+$zK1{0Zgy\p%/>ybm|oo<߿?٨גz@^K7+YG[Gz@Mdn%hUYNh||}%Om V M5 M3ȯgn=h|:>IkɇG| "K+" Dk7r+o0|>iƷ ݀VۘϦ= f|R ba;*&焯|k6pR .×_>|]`+ M3W1 *f911m1[C-r6++rUN4;E>+|VNy81_/O/ |6leZyr +[F%Zh~0HZx_<ʣ9[F=Z;|c/:> kɟ7+'J×/WD~)|W[W_9| ҭ؀*4 E'z@L%Zd/K6M5"/|a+ MS!UhT5ٍ͉DF$E-\ q/.jjB>+|VN_t|y|Yg/KU_J4*Ѫ@܀OkS_‰8'VN8%z֣Q4_w*wA-q/j˗+"RJ+-߀OkhF;$r+o0|qH  hBs_t|_Ž;bVvH9˃/< &KJ+2k_*4КD{QH`HvHόϥvV]C$u,a ;ŕ"wo;qc6]:lJ`ぱX-0>|.8+730 +c.U=hH`ܟ9J]VƩcqLUw#xN_?lYrE+jѷcH"_BG.$I훓eD %$m$ %M`y}|QN-͌臨DQrM˗`38l 6#:MwmvBS1wКfBQ>[x>Y>>gWDlmQjs擙;hDsa_0ͱ66Gj>i"?hB"GcQ щq~F|E6:᫑[MR·h^ -d9,'o|V>i׍QV748"3>Z>#"b{\Oo'|uWNu-[_ |5>i7F9՘Oi Φٟ_ͯΦV |BӌKw} [Lvcy:: >>luy|3ɟhLFk& D/P%M-q/nj[_Oen*hмm/:c|__<ᵎij$V-o55|(_<̓o|<ʣ9_1-@b4x E'L%gψ%ZhEo'|uWNjુ5- MS!?sh1\JK._r\%Zj%ל|7l1㫓ۅ>]iI$2s'1h3@a//^ŋxOh ME#-{ EwLktb|/1 _j eFƀOO_|y̓oZ4 [FZh5H'LO-/~joHo'|uWN>2[&jુo l4|B_hGk(cD?2[xXY̭_؞'6Xk,ǼܗxmY_fȗ(Ɍ/IKے'6X(I˗fb.b r _Oh"s7Uh.6D;"f1\jgs V ̀obn8__ˇ/.j [FZ3МۘOi6ȡ97I+rUN4;E>+|VNy81_/O/ |6leZyr +[F%Zh^0HZx_<ʣ9[F=Z;|c/:> O7I'J×/WD~)|W[ W_9| ҭ؀*4 E'L%Zd/K6M5"/|a+ MS!UhT5ٍ͉D/P$E-\ q/.jjB>+|VN_t|y|Yg/KU_J4*Ѫ@܀OS_‰8'VN8%z֣Q4_w*wA-q/j˗+"RJ+-߀OiF;$r+o0|qH  hBs_t|_Ž;bVvH9˃/< &KJ+2k_*4КD{Q=^cFܟV"E2β'ej-4fgn;x Įv8.Rt^vvV_ 5r|Du:U}ߪ:#g v'-TW]Bkq MvKv7@&5P 5qYnǨdwl{yľ(=ϴV|d?#~{U:sNFGmkX|m`[vT Ǽ1Ns#?OK=,zu3Wg+$f:Ekz@K%#e? /$|qv(jTѣOP=7FQP-B= |$ѧ=i'~*V%%*?$r"'qo:8߹k%KIOo0ձa>\Ҡ@'xUfZYqGw20*"+W?:ޭޕDG1wz%wW`9Iksmتg<Bxb o-VU7kɟws{E&dCc?mQ6u{݃LA=&dzҽ빝䥟EooqN'ş~z7C/D/T|fʥ;A/}+[I"Ke~^ڣ_wl;mT:*+6J+VβX*_ꡎrU}s(ku=HWqzOY{j[yP'3a=1x]Yf;?ݏ978p:凂zP߸ضZs(uh_.+ [lYTb1 ij61BY,wM%f X^N|7f=m=x [wc̽ к*{ e/mƗ랭\J#FQDv[Oh}w/`6曫㫕ۉ^·jJ=H$NBcZ9׀ud8_Hs"2("zFR>ig'?h$eGudc:#|_D/Q uD÷Qo%K[Vр(DIKo.|sE߁VW42_rDSh%_U_$ޅZj{|cu"mo#|a[B~?CZCm/1:jOK Kw# "c EW+o/|;a͕{I̝$Ds_XJK,_bX%Zj%֜A~2sHFˁfX7:qKo^눆oJ%Dsuq:8j\73Zi%N_ht@+/4fK3j w V.l{ 4j6·mOh ~h AsL|Q:xj/Kİ0yf|dhi8K%h=C NŇm7ܵ*~1ŵ!n>i _2|% ۩|:|q _+:h*$$s'1 ;Ѭ0{Y8ߒf|5S\5V?#hAh׬1_/YϊΩ#JI&|_o4 VŅM3竌UW_ekf|"+s?*1S'ΟM%Z/K,Kb5"u|m`s= R+RiKXh>k/1|1C-b%U /d዆T:HB >'|`᫐ໃI̝$@€hDUE/w͈^@ZD/KW}$_:i7#R/|W_| MS! sќi/VX%J-/VjeU_K/da;e5_ V8s焯l4||w?ИhV KDPDH G{#h" E\ϊ) b_, \W$כ~|,2[eE_, *૴P+_,|/_Z4jz U/M戢 6u|kӀo7UO|=o4S5|B_X%h,Aq4g( |UOG|mW+= B9=jWc/ed.O4ғԎn;y90?zgzw:?̘`i?OK;~P퇺Wuyw?uwվuӂfRORǕӞϨSmS_j-[뷪ouGƈJۢLV|ZP}&i|^Oz =,~ZD/M=V{W*?t~z uZ:oR1%&}3i{sFZb9IKN={APɞ!F?dͤ1^,x"=|9lr2= %->A\ь19Jّ{v$}OsM{ Zj᫅%Vjh0wC$[c]CϚL|N(95)j'4"/47oo^twvh"sg1*42OlwG2HPrHV |Bӌ/|sh8C3?1__T HPrRsNij%^$%|+[ ߋ,%oF9\)o.|3R_)E-BR4Ҙﰎϑ*/z:~L|N/*U/|o|;ેߤTH_-|l+5|B_!h Ak4R %M/iF-4% IViI3糒 Sϊf4;:; ak=%WE @cZUh>e!]/N-t%IV |B_w0ׁ<4;_T_2 %Zij%^$|[ߋ,C/S=SͅoF_2$"QV)?hwDh!/-E iAZ4ેo|;_=lG &[%|W K MS!s1hNja/:ɔLjd/K&K SϊfL;:G;k=%WE @cZUh>eסe8_EK iI-Z/-Ki)/;@ÁVZ_T_,%Zij%6 Uol|GzR'Gi9E2OTV'pf Fșj6y]:y\AOh hGtH4j/iKJv𵃯|~2 E: N][Oh O2I4Dk %U/J-R{EJR5*/%_|؎_.:e— _.lm5|%F?(4𽝢 I)K P&sqhb/: $sܛ("IVN|%W_|%̀d$K |@no|_%݌f֣L4Ԣ9i4V5"燯-|mk Liu"Nu|=a+ MS!I>ƓhAsZ|tb/6ņبOh | %mk+N&|e— [[ @ Qgx4ƣ5 ́|o'M2-HMIk4$/`4ͷ*z7$7}Zf*όϥ_~NIW֪\gfA3շMRP[SVwī'WۃW8xjjIի.oUƫ4U.Y[uEu:޽B}&OέD\33 q<uDxXnߦvv޸$F${Цŏ+ZzK p;1LkVlgvQlE=Bu t͠-_WemF+Zn8͹K-r\%ɥVɕF"Zhs EwH$:/_ uH÷^E-o|h7++/^jߣ= ߃^jW_ohG%^%$G-|/$_$ކo |[۰1߳o=|{E>i7{375Q>oO-wwS+_!wY".:{ޅmQ iO`4&5 G ^DCGbim^?oħ,o5B_h"/?]ka#>eS0kLVF\cȷ7k7λ;|w;|㈩^-Zf澅[h}k~>* _rSڅ*%hoqFsTa#h{H7Oho ߠ/4?5QǗ#NqhiVF gp45 6yMR CDbwZy[mw\sR܇h?rUn%|W*.1{"7o/_6\'4&澀 h=t|mˡ9m9*/N|9_ EKK+Oh 3a4Fk*r ţx/C<H 04A_h~iw/G/ 84Ў5x |h Gk7mr=7p/nō[/n !~h&ɖljd/K6U|= K /_6\'4M/4OdoK_\…_ rˁ/sEKK+Oh=@}zq.qiηY!b"1ڻIqJ 04A_h~4__:qhơkZ9$?4o0C_8V8$sw 707vscvja_}Z5"w|=_H7 a  4_` hh1YkQRIT|D)@bgcP{Lv뮌:} i8zzoཚ2Ԯཞ{B^^ӤC:]#ńky3s]oXr6 NC\SD\7gO]kyRzǃڛso)AFQP-B :$[tBzo'"Bݺ Q޼٭B~5;$kt~$zyy?:_HyR}eK3ͣ^imU/\;q^){TOzڬgRbN, Uwo[$o~xT|M{8Bg[MOY0 z.'^z?k꽹nȳޛ}M<WG/죧N}uÛ^\E=yىw[&b|}NŪmQv lcgtEli]-i߷XT#K˶ȢqE`](kݙd%b}@XrK*{h{_ŵî^ŋ̋jqurI`{|mƥryl)/&0A{!&޳! w|uǾ(=ؖ.?7>X,@/}P tZlg|o'CP_Jn yowt?\_E~zz%5ul?z ןf{logs6ͦ7'wCmwVLp8cN=&\}2S<1#/LZn̕KN6,U"oʺ+MċY?xiɫv׸Ǿ8zY !Q&/M[Q<1yؖ$M,>v7GWZ㮴uh#z?L\\xX=CyޠG:G\]#ˊ*5',}s#e"MěoK|YG,!E7,t"ByQftө:OtG҈ndyZ{Dy}lATK|3rb_LbB{[=r[{&zO?ؖbiwtǜYg x=0O=[{b}=qz3Sx-CPzRDxs޼E?LycήnCz(|^tg22.^A~n:VGˡO|V$*6cvY^^]|v3 KPr[XyƖw-0_wOs/H+>9^".Aj,Y7Y $6Lݘ.!gbz1cyҋOng"ZCOڼ)_'>au{R?w13K\KyqU<=/"_ _x^!dUD\v!g`88-_^^FeO~L(h/xW_{YBoG{hwGiʮSOxjܩ-5zq"G[bܛJr^ 2.-HQm<4notļGZn'5(^@5[ Ͽ ]-' S$֝ MK6*mC}CV]_~l3igo9/_(w Vx{?|E9Y9uU;9ԿX~]š/~"b`"Ȳ߈Oy1۔3R>ߨKݟXƷ)NOٟ54z^Gr^{>/Fq>+cH'0Gr^{g`8r4]e0:>⼶y> CWVr^{0x^=g3j^r]YiMK/z|â]X'/yhĹ׏K泋+VbNJC_(a+W$y+Y)ڧU"aFtZ񳯴 mo q;W݃?%"_=FE֋Q\HazeI*m妎M6ڵX'{elgID_ ?\a{YqQ^y1tļy؅ςa[-Va _8a_+a}\BߏԞŋF+cg ϪgVk?Mbɷ8ztMzfui+V1ws(!޴1}PbS<9ԯθ EŹѢљ;ND,0n2MwFzG{],7^Xvme7^gXvgzMS SS (%y1tnbޏrg`*11y>{DZ)b;cʊi) ?ܡQo[~I92+TucPo/c<<￟ ‡L"}bQ4}l1Fǒ9hwp>6}r{Qݧ\v̓)[-y{B'#uY@E^ {[hCYюwy]}şfbyuz}q wŬx)ׂY8\lgm?^H2}64^}w}[Ͼ{{~2-5xھǢVu첟sO:q3c#>q zg-284jZ!^GʏDOU=cq]aMq{>+{E9`h5x#}rqo;7;8̝O*Yї32С30p5wNl >$h$ vR3670zYmԻu|u`_^5{Ssz]1׬[Gz>u^1T~ ^Shqsv:ݦ^3hu{:XML_9%$sRqU鞫nʩVcקּ*CՐ/${>o\b^wp5w V>:U>]3vڿY >'Qq1Z|cٞ{8qv~8{ 8CPlۢcmh:Cmh9E9{Z4:F7uog kM_wVː[, DFAP)Ksmhв򭖼N|/>5;QL\`4ES-o1Za[AhWZwoߌ[,UUW B-_i_g^eh]Vx+˷ZJ 5|Mifym|b95˦uM&?#=hN#|w?-_s#.1XN-9bSa>iO\ x4[F4;q4}8kUK- 6sg1hVu󍴚MR |b9pZY_[}C- 4K¤Hh"ޖ{uo|ߠp9|૆j6G'4n"s{ ٘︎e> rbma(jw\| nv܀oZ_$^o;|k MS!$sG1{Мe/:hjD/K4K0WACA+͖-whG8qڭ᫖{9̝DڀS_b1K j !\>ڢ)1vXi/"_bXhbK :Wan7(V/|૆j6G'4M斠QMh5/N_Љ_8%ZiE| nv܀7q:{ mo;|N'4MHA7q,4fK3jfB\ x4[63n CG8yCn _ߓa4$|u//"~>_U_%|s}hjfxY$%Z$/ KВ W[:*l %Eo|UW "h_%-A24%ϗhwYÅˉ%RDHo7| ؎&Q/|m5i?#=h&1afk@É7LUhMUL8n"s$Mi814AÉن|3W N&+&+H&IohSD2˄/DB1QɔRyv))oDNCcZ 'N F4AxlDD[H ]Lb5W|K!$ Ie_"|eZ> __ dH4F5;|[$MMO$N^1>Sꯘ'4!7swwܒИ"y<$N^q0S+ _; .W_ |]`kg0>od9@ od|4[A:CcZ+\ߘ/Q'_jF-E'o5V%_l|㛅Hu|k_GJ4|Bӌo:s;9hMGs||t'/q"_8j8 !_OO$ ED'D||ܑhDk0} 6%V_bE,/*Oh hX)|P%_bB~ |%v| ct#//|}H 똻uh@sa_L%ZDG'DShHe, G_; MS!s1h6(DIK_(%ZEi%ʜ | | |'<e/:>_ tK/OWo0#;h FHDJK$_"H%ZEj%R_hF47GOJ-/Vh4|$: ^3l Zub5__j*/c:4֡ͅ| :>DP%_"UD__|>|˂-o|_$:v𵃯#l4|B_ȟhFk:# ŢX/B,H | | |'w|WkLHDc$ZSt~_mzi kl ~YOTKVޤst SxPWȭܩ#򭁮!wkym6n4cYc奁ԱAX{}z` e'ձG`XtKpuyz(SxOVYy6c֟: ^}P}D$fhUK{ЪL=^ӿczb <4XD/w[rɼľݢhm5=nqbts%E\\QQYq/6oR? k89-0κѫk:ƲW1_\_\_TZvz`,|z>e::UUi: ۬.;kU_nn͚3)8+=nUpsjH\}*Sa'FgWT|@;=gIс^(jEhc992O|fr{;ZDLED{',4e?;o,qNf؝3#>" ?=/z #\mj:ߣ_WOVگ{{^]/W PǩwyݯiU|Wie5gS]onR>yyVjl>YgOV'}cF1Q{Ѫg:&wh1Nӷ_K/w+{ɟuiAZpݽ_M)WwgYs:^;il2 ZhΧ'%ޓKC+51~'7pO Cb#ɝǼQƉ㘋VGk@/\|<~{U~1֯KI-U>eTgctܩ-Mc%X\\[N'rD]kqf0 cKQ\bظ`jر֓B_'/K<+*wۼ;r]O0:'#I;}鵁;6<*rv>N{_ۗobtsk=U',_;&1>/n윿MOvw073mh|VؿkZ8m:Wvd\wY۷Υzt枒e:weY}Y(W^F _,3žư*?}x~En'+._6~u|Ҁ ,\Z9u:륫/ws0pYyd+0-?|HK/tM)Mv>(vyW|*=/)ƋQ|c{KnyٳFK@]ْ%˘}[Ĉ-9g- gciYaA&f q}o <`Gw7,5*lIvg}wbz:-[FȋYڿ[ߛmYcbh)yMnQ5Di^j"4Fq &c'KrCij<ɢz(p~k/dD1/`ف/_ȵ{cEu <na>/<LcGGmpgSi-bn_[#%^<|Osa+[~׺cfؑq?^ #ۮO}gߘbNbWC*i~(UnWsrrY>}#zr8"bPѺض7~nNJ+ſIc*ě˿P(>bnTp݅kf+8U ;b?[~,'+L|'ʋ:Z#i|>Bgjztv?P^mwN0caԲaR|ARڜfHݜfLΞ QD/EZlLKǙ wNA\RZ"ZbHOtx[bc3'9.&|]|Y\L*|V|3E*^>krkszZ,_?7 o$ުx/gy$c֭ԏ]~w֟Hz6͜{W-3fFNd|,ms^UX{*RKHI2oC}?@(W^Z 9zdA}nױvRhJ">h s+zh`7D֙Vo }(0k^R{%^m·d!s.i!YwJs>4?յ)?Tx&~iiR%򕦡FٕFnYWfӧ:_>o*_vo7viS_^o6N4hM7#l*+ 061!mBD4 `P |w |ZA M TH s4ZYb*3DMf ZLAC+R1g|5v|kNHHSn!h=2sG>%oQd  -H3&Y ~K;i]]]ґJw:k_+N-*èFZY7ͷ-H3v>Tsu0B֎|26 |IhH Ie6 _(_Հo; 3|kW*Uo |7뇆ZYёUgymACi|׆IE_|-`kU+k_#l5>T/(B4Bze6bbG[_/v;./vuo6 Ew :k_+N-*èFZY@[8_/G卨kZ.hf9M_mq¿8_h+ _(_|m[_S_ W*_)/N 뇆ZYw^gti/. q\|5/ (ky2W<_DN5| 4U ʇP7fKeŭn7ōrK[_ofp+_T)y)E%|UV|uCk54h8_O-|/>m |-/ (Zv^/<>ʛU} S}Ɇ"r}oEͮ4}E(V!|_".}#ۥW.CWTT|NS `V.kVP׀TTH%U4 ԝК m{ju/zv>TkB(FV4<2>j|4Nxju׵y<|~eͣW i+_|AW46|u#Ј@k4+<]u- HHպwh+w;/ |Fg_bw<__&s L%|ƷèFZYwqB[juoVv>T㻀m6h] WGϨOlCm%|A |SQ>!h5Az̿89N'ZȉrJSA(FV42>S/q|n |>?Vp_|` IHSտ|u#Ј@k4+ EGTmaÿ_h+ĿhK|F6A./2|ne/#MUa C# ,+[<_M-l/6m&/6 P mк6"3jCGmc[F _Pn!hMfP>C_24fݳ HH/h _P>QhD "ϪŊ¿X_/VU_4e>?[*/ _|A-$#M5PukQYaU/2>zƷςX_,պ}E_4>#KE_4e >7_GèFZEeŬf3Ō2KYA+{Wu' EgMǶ)(B4BК͠S<_I{WjuLuuv>T/(E(4jS;|q*8>?Cн9b1󡻱󳱴K^`%/Kr?$~i3f_,_ 9=ޏWC}E%JNcD_p2n_J_5So/Op̖vrv6_ީ^jic3w'IZ;y'O$sXϾ;x+;fca,۸s7/υAݟ}H]7xNAC|ϰmRtMF<`瀼++2Y>\);&i4׏:v{uz߫߉pIwXDƉ?~;7àwM'ؾ-YI;הw7D5<)rP^Djwv'񽃄U}DyK _Ref531761766}DyK _Ref531762011DdIp  C ,Agraphy_lfC"b rbԙ= O Ǩnrbԙ= O PNG  IHDR 'sRGBPLTE?;~I|IDATxM*3xTrd R@1"_“_AAZD#x~iFF>EpQ(8(tjXn C>?c?K,:F _MgZQW5_j.Eۋý >>EI&Mz^Э}پJ&=FprQG!EO# JOUiM tRsxT#pNj d Ʋ )d" I*cЫ0ՈFЮg$7T=c{Bk{G5LJvyPmqY>~ɀʆ–7hBsQ. TʋBOi _`]M>퐷[lCn@6 m({~z{c6m*@o6J?ll`3"}P)r[(I Js+ɓѐZRr!mY$K:bdA&2I >K#]$FSH&!cSAcWWPSDKI6m(]ľ3RH]S8b!!R(ϩ: )4C,pK)e[52/Zޱv6E=#_;( z5DYR>裸98Խ(;%m((P19D״ bE* c cYßa^ Nh? ;3Qaf[q̏Ta@GJ;ҮQ?C02(Jǟa5ͫRXҥ0v ˀ 7ʵIRv~(#\n~(ђDn~(C!ʹ(ucM !"ʒ[1\ fs(Mf~`}U_(弻6{S|>ntAW bhOUAR6M&V PKv0o!)0SܛR߮}@ {,5 Ȃ8eZ(|f= Ѥ59QP=x)=--݁4޹'T.) :54/DUqrRnJ[<&:.$)PCRY/S@!Yn!$s Fj02S/Oe-tbpҤK2Eq2[Pk-z\#gQg^ԩ!_Ðg&J )6 "V:wHS z.i/aʂgwPD)FXzGaȤY,qhao e_t n5<]| $!U VZTJGD!+7x8h8C9D!?-)!?X8 x;z Ϻ:lJ p?[C uEY%AԲP#+BvajY@dB!he7Ls`F- 2U[S CI~DD5.:D\'")|fZH `k}RpY|s`R|C%S;`_8q.\Z97Zjoq̞b_)[(q0\J]%vsƁ{%P-i;o$m|^E<:xK G~Oԩ8¬NVX bT uױU\Nڎh;UB[.%Vks׫֩8|S7 T^Nņٞf0 21Tj֩؎8# aӈJԫ*/k /)̌,I$)W>fر.tUZ"p\Z# FI't;VS@ ( C-fz_/gc?})~k)( V7>Wbo °m݊0YH!Tʥu+В+)aq Su+tY֭P\z0YN!p,h݊FTl؎bWQ֭ؑ KVٺGPS/.t-H|_@܏ޕBO;D[q rK +\uBt<HZ!$I:k@Wj_]}a֭hK<e]sC ޺-}<_Ls/ /m8) GKo݊IC˾sj{-h6  uS“QxgA@lOխ5?j 8[#~HFY|dFe˱“91i,`?i6T< #d;tM<AVQpk%$#C`9Ǩgެg#wY_2 swy͡ܖR@[ݲ YYQU x^{Qg9Q@czsm P28!.L(@@=(ϫ׳ KPZJ=GU{,Նep_28!*qg@6JY?ĤB H~)e2 K%3/ ]yPNwyl*.: N!hJs6 ks%N6ԣSЦZyROk W= v}!pߒ x=) oæO(2kXN;bUM! 6 b Xa~yP@@?ρy>SxbPx309GUkx ) bS|}&af&64~[qw}Gm1bowx~# )7*q)QG}~7% 6+!yx;1AUOĀZg1c#b;rȉ5Oz~iTR㚘$$G#IuKGeCĮ0*iÙf}>Oj{yL{ϣ~ݙ_L5xb[2!Yi[z,ں{]bUbUȶWuT$6lhZ<9{ yJFuڼ-MS^w%(1nR'>icbԣ8MM|V7j6Thjn vXwX-lmoo"߀,MJ^Nijzlfh QN$p\dla^#c}q}XIoqJm-3*1{v/NS3 G<7p<ì]5镮#spw=Bun_ɬ9>-M ?"ekĮ*nFbUGLh[Gx[l^ݪm{X)a<+9 [y.{ m޶ {*3|oĸZ.A| ^6oGX }"&0z06.nkhcz-l 1 5 &0\IbOs=J7C);oG쵚K},zM@i$ǶQ9G%M^Cmh&6 €_b&fkɢzX vJlŰ:,W&VMJ,9??,w]LdL=#~2:>oKcQmLFkN]3d~frլ bRm,6p!C7 j6-8W/cyɐmy"Ý@ &Lb#&wJ? &S>M\,.ј$ bbk ў2N'sbWOmG'/8+إwi]r1o~n8XDVb%FĠ;B0X "j#j#c}%                          ! " # $ % & ' ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 :  l = > ? @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z [ \ ] ^ _ ` a b c d e f g h i j  & o p q r s t u v w x y z { | } ~  O8s} # 1a4(Mm#*EsAoj_c7`5_R DǠ1irA!@q}鄈<&}#&۰띱5Qz,ȯ.^teJlo,.;,!6VbLIYXVZٺ:륙ĖS%mlݤ꾷Sч$64 "Id¬[poƱpQ-B)V(܁o7ul3NцXimbO/Elg.ilnC ck'[3ʆx#^,O>3MSĬ?BL{El zxoD]<ƝyL{ͮFL_x_FlϬ[vA1!VF;Xq[!'D%DirOo~Da1W>ԃ̪{+=]ICA&ӧTQ|'&b4B,@㝪Z%U}y'vNo 1j1z\|+m?Pыd}5E1DŽ1q8M_&2CUNG vF(ˌئEC$1KM(U5 TCˡ{ WhTv,n;CeT&`_k :{`zǚ_&2CUᏧC J>ʏX@+c9Ub|o4b[%vIbYm֓;S3;ؘXN>23+҉~&@O  O)lyCXňŪ@e1Leω)wR bPkc.SHT ZifvM4pd@LJx}.-;rqc@}l'؃=eN"?JnLmlbbyKB1rƣr'X˚ <&$2Xa_f$14N+rKl̐ѭbGU /DL̹Ja˰,cK03o`zԧb7',3ä6^9h>jgSEK9>X>n_sE/4F(L|YA^r'p>3 '{$~ vvV1]gcަ s&*&&ig[cf>3ĄvrscԟX|TMDl2fODSb 8㓷'gwSљgfR# v`YQĬ~6)V:`6dƟ$UbXjb*ČR{% Բ5١ѨQ[9POILs'Bt&<ӿӡ:}1r??9Xypׁ1rkˈ*9g7H"SC[9[3eJ]J9dn,;fy-zx'Abtlc7Iʌʕ֥cSGB~Gg_ 슱<)J61+1@s}LˣS( 1wTrQiQ"1ܙ?_X6#`Ȫ*~r{+N]3ֺ`柟|Ğvޱp~3ğf#sR{c|Lb~_*}JۢwץWk]o~P')%-z7leu+7] vZ9{KMl޵Vb&cbk.oһ\%)1+kb$׫7n[@kQ!ض7nʍ[M+ԣkluc^]dB׈*,]kb%VYJ0['<*+UYJ HQ*\%%&򙭭H(dkLVYd!ʢީL,>~LLlUESۀ'Lj*"\eq@"ƻRr;{NUeC'1UU>=bQy7]U b,7̛u+[SeQVWcdyLWY1ن]p;=.{Hlɔ#v: +>X6.;,+~Į:,SkKٯá@3S15HߖW2{"F *V\b%%Vdb|fJ҂lUkfhgπȫ'\3ё f^3?^J~6h 9>5{`C6W8Ebj90' xĨ|_ز \@diDIl4.h;D>*1.3+b<8*ǀ׌_ib"VCYQ:.߬ÂҜe~-b«رa[^2ĊS|R[  aYGa1WM-V=9Rt}fvYeNOݛLbT3+v&41u,*o:W(p@6z͚N: +$$sAŹe{ bAc^V~ZB5U,<&4ьÔ~i2Bfr/#?oVYL0m&Hct_Q,<Ǹ"];`}Feβ^v6IkiLF%jeab"]V˚`|<&$2Xa_f#1m}6-9vӯOV a9uT [ILe%Q.R;O1K|fx':Y4g&֏o':I} ˝QQ׋&vL 5׊NJgAWtk+UbX%VUb;ĜT{bhS_@ Т,|5Sש0DzلcSIz` 1nc1}Uxǔ6=%1sdb*ISBPq*Dʬ4K_c8*㷁t.rEf~Z]0;JL_:.Ol D^_&b]\z*I} ׮\+5U:|-g#}lwo߅ݹɣ=Q}Cښ5*[tHwDŽy 98w ƨm@b<1.A*FZ߯~69.Pɘ4I M(Gla+hTzoOQ|=Dܞ+mTҌژO Ra*!9UG'[$"WنYD~bٯ*nFb*ļ_V7K vTz bPK2;bAF,Y &D셅ZtJ+Ckb,p^܂|vD̛_/N5ok8oBYe&IF:pCPCwT[LX?yIڕ>#s޹?2DYCx|5Hj>=bQy7(1EE3#uE WKU '1E0ҞDŽfK&#Z7٣X` k>oؗ}hyObUŧKNJ{b~X 6,ՑY(^f_ql'ZƖ"N+ 7˝;F2Ī"&*_uo#9H󈅫"!l1oV"1o'/VEgxS/߬Hb-U)񪈘O b4CzYcXUDc[C,7+h !,^Q˞~"رm峺,ClC˞y:XztWLǮlZ"6"vMCudfTFvk,Êk%)חa5["La2bܤ/Nafɕm$v y7I &C}wrs},1`W:\?kF!1cӣrkHli%)b̜?;k.@UCc~3q%ezfY&^d[Lm,CY*b˩;fwصZ2bU|6[u*̓rXUeRb\e ܔtXL,iCQ ?4$'CXLbԿXJkT53ʨlږ#g˵-Y^c0q2خ$mY"1nyLh)Ӥe2.s"5CZ-!/`$1KM(CX,.3XLb^b8zEĂ-$@9d6=O0cfA/qx cpy-$vlc̊"JzAVEH[+3[ClLFDٹJLL^ nDT?լylo˴}ACXňLM0tXێ&ym ocg ̪9^̤1P>H'C?FxVfҼ1N}ό=F2ѷq:H pTb9-É4&c01IY34WL;?cҼ]mBlȒ>}_ ĊS|N'V\'[Gl;6,s Fj*=\lVv^C'oRi#OX+t5}D=܉_; ; g;/g@% :;{s`JR8~T\Bv.ib h>5.Rr9;:bo")')[/eט)5Ooq!\bVkyVBӭ.~Fi}髉/=X%VJl*[8u{i>}e k-&Tk fwߜx?ro & نm9Pt&qb3C[XČw??I)8pff- #9hYZFcू]ïbF8QGӄ/zAKF~ \} M~Ć*4YD*+ru)b@7MT0<=feTK"ƭ>}bRX{>o1I,>fL:Qi&sG8E>5D,QiJVf`@cn"-"3$*ie?@Ħms9c=Sb YVUbX[]D>.qYhZJzS"TX>Ę Ɉ^"/ڥHr #"ΫXJ1b:bI!L/?9:["vN-T4CUx[YX&{!V[cΤ1Y@L~EFegcb˳ިCjߘduE=>y$Ϥbbb~#˰b2bs}Ұ/r⏈ѶXU|*Ēra)š|a!񗝿N쵨˼=@_V"֐6֨1&|a|)Xy WFpJR0_V"1ߕ 2cˊ$R&_XP/sX_V213!/lDbY664U‚zs_Vd;L,ElCkcb8إb>ve Pb%VYk{$}v"l=v>P}7 mhX>+X;V)gv.}f+gv21y%Q P=#u.mo˴}ACXň.&D: .E @j.20P3{Xϊ'>3i ;Cgb1tdq"ƩqRG+NN1eoI2uˑwM'?Sc4ZQ5.H˿Dl˱mYU,ӯHv:5:@֮7,+ ڃJO`&;eeϋU߱8I<\# ;K|bٹRv.A --Bw^ ΀K+- +a"`k@S;q-5ٹX "=( X8;תQ+GrvuDSb O8S'&;W8Rif)P.1_< fe?Y]%4߾KXOl}J*t-:8 h1ԢigeoXOLf ;rdh} -GbyVN9YOa45 Ykqбd1=3.g;Yrl"F9q,uDٳryz |- 6BZhdb@:8`[1m2(%!&3Nd5ۼ$QiDOrPyE!bJYQʌ h8@3Ե Gl66 6Xc;%bJ*J/^~ R'6'巜3Zsr X\Q`ćsPV,n2Wk y^\^P WDbca ºF?d?#=7aiW(Ϗ' dŢ>@EM1uX;dga{y}1nv~uQ~Xf;E@,{rm/3*mbloV,E,=68Ã{ ~t Bٰ߫]ORq0+|4=4V_"Vz*J**J+|wY0~RIENDB`}DyK _Ref530883055}DyK _Ref530884156}DyK _Ref530885349}DyK _Ref531761766}DyK _Ref531402157w$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655Tw$$If!vh5555#v#v#v:Vl t655T}DyK _Ref531402157}DyK _Ref531403222o$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh57555#v7#v#v:Vl t65To$$If!vh57555#v7#v#v:Vl t65To$$If!vh57555#v7#v#v:Vl t65To$$If!vh57555#v7#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65To$$If!vh5555#v#v#v:Vl t65T}DyK _Ref531421544o$$If!vh5h555#vh#v#v:Vl t65To$$If!vh5h555#vh#v#v:Vl t65To$$If!vh5h555#vh#v#v:Vl t65T}DyK _Ref531488842o$$If!vh5h5; 5; 5#vh#v; #v:Vl t65To$$If!vh5h5; 5; 5#vh#v; #v:Vl t65To$$If!vh5h5; 5; 5#vh#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65T}DyK _Ref530816124o$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh55; 5; 5#v#v; #v:Vl t65To$$If!vh5h555#vh#v#v:Vl t65To$$If!vh5h555#vh#v#v:Vl t65To$$If!vh5h555#vh#v#v:Vl t65T}DyK _Ref531501615o$$If!vh5h5K5K5#vh#vK#v:Vl t65To$$If!vh5h5K5K5#vh#vK#v:Vl t65To$$If!vh5h5K5K5#vh#vK#v:Vl t65Tr$$If!vh55w 5w 5#v#vw #v:Vl4 t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65To$$If!vh55w 5w 5#v#vw #v:Vl t65T}DyK _Ref531504567r$$If!vh5\555#v\#v#v:Vl4 t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh5\555#v\#v#v:Vl t65To$$If!vh55K5K5#v#vK#v:Vl t65To$$If!vh55K5K5#v#vK#v:Vl t65To$$If!vh55K5K5#v#vK#v:Vl t65To$$If!vh55K5K5#v#vK#v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65To$$If!vh55g 5g 5#v#vg #v:Vl t65T}DyK _Ref531665805}DyK _Ref531665809}DyK _Ref531573279}DyK _Ref536603665 Dd x  C 4AaftermovementC"b n_Dl; :n n_Dl;PNG  IHDRsRGBPLTEg&T *IDATx[*@s?N k9?+Cacm$vr) ~?K[vՉvy߰{|w9/i/@og׋Hn6x,^BS`'?vzYO+Nc,$aAI_v ,QL0Q^u`5% 4 0.{uؕdo; a5iZJ1إZ(Uy,VN )ؠdx@ː)o}Xn ցk7}a[z\zO[߿8,ܫ՝ y+"]X 5MsUs;%aRaW>^I,.~:nV1(\c_m4$\۫. ,/\+F¢5Sg63e`*]+xװ TfOդ}.r݃EU-;("d,~$Dzamd:]Y^kvLU٫#3>"V ="`__{%`Eu mq-9G?~{8o'ō9-C}X[$cctkw ~D}XWy#D&,zXsw$˴@~Y_Ia'&ڔ[ ,Dؤg+,w_VrCz ,'U:P.| G?%H¶".Bׅ>$]PcF{jXaET$`m%IXQ;J+c*0t(~e_<ܩ3JWn+5Vbn&6C6D[e^Iojz { frkA{ ߞgϪ7ڗ&-WQ ֪ocHfm#Ӟ#g^ܱq[ Y{vٳUa}jU%oۣCLgm"k*` +Aq`=;gr*d9σEsNv`~o? kهfy{6r֬gG| }`LDNOoҰ<XspKwFGxبnh?+.]%mh-=z_!lh2zgw& E`OЗNwC3o۫XeY>n+p#Pu&~K,vtA.u<~8?ݭNP y|\ ;d öOuuvpd#;IaAJ1sO>m[ $\vqdWi jgxKm&lzէ5m* ;tP\|p:(]֣qruYO([WaX9ce3ڰ[]J;<ӐOle gEPRINT(*< ]CompObjhqObjInfo)-jldOH EMF]A@F, EMF+@``FxEMF+0@?@ @ @X@c @$?4~@e&8DMC!b $$=='% % Lda! .!??% % $$AAFEMF+@<0Ne>@H<?K"C8DK"C8D4~@?4~@?K"C@$$==_888% % V0d%).%).!!%% % $$AA( F4(EMF+ @$&C.AhfBbɛC$$==% % LdLTy!??% % $$AAFEMF+@<0Ne>@H<&CR\C)CR\C)C.A&C.A&CR\C@$$==_888% % V0KVLLL% % $$AA( FEMF+*@$33B%B&C.A@0$9=ARIAL6@|pContext;>I>9>I>DZ>I>UU>I>#>I>>I>r ?I>??   RpArialMonotype:Arial Regular:Vek%` 6` ` C` `N`B` '<`<` `` @ ``xɒ3x)wg1h g1dv% Tx_"0AA_.L\Context% F@4EMF++@ @$iC `B03KB#B( $$=='% % LdY8,!??% % $$AAFEMF+@<0Ne>@H<iCU$C&CU$C&C `BiC `BiCU$C@$$==_888% % V0W7J J J % % $$AA( FEMF+*@$33B%BiC `B@0$9=ARIAL6@pdParentǰ=?G$>?9c>?丄>?r\>?>???   % TpbftAAbrLXParent% F@4EMF++@ @$3iCZC03KB$B( $$=='3% % LdYF ,!??% % $$AAFEMF+@<0Ne>@H<iCXC&CXC&CZCiCZCiCXC@$$==_888% % V0WIss  s% % $$AA( FEMF+*@$33B%BiCZC@0$9=ARIAL6@\PHead=?*J>?丄>?r\>???   % Tde ~AAeLTHead% F@4EMF++@ @$&PCZC03KB$B( $$=='% % Ld >F -!??% % $$AAFEMF+@<0Ne>@H<&PCXCCXCCZC&PCZC&PCXC@$$==_888% % V0 AIss  s% % $$AA( FEMF+*@$33B%B&PCZC@0$9=ARIAL6@h\</Pr>=?9N7>?V>?g>?X>???   % Tl 3AALX</Pr>% F@4EMF++@ @$jAZC03KB$B$$==% % Ld@F -!??% % $$AAFEMF+@<0Ne>@H<jAXC2ڂBXC2ڂBZCjAZCjAXC@$$==_888% % V0 CIss s% % $$AA( FEMF+*@$33B%BjAZC@0$9=ARIAL6@\P<Pr>>?rG>?>?q>???   % Td 4AALT<Pr>% F@4EMF++@ @$@BZC23KB$B( $$=='% % LdNF -!??% % $$AAFEMF+@<0Ne>@H<@BXCCXCCZC@BZC@BXC@$$==_888% % V0MIss  s% % $$AA( FEMF+*@$33B%B@BZC@0$9=ARIAL6@H<SR0>L>r|>L>??   % TX`oAA`LPSR% FEMF++@ *@$33B%B@BZC6@\PPre->>j`>>'>>ʢ>>??   % Td]qAA] LTPre-% FEMF++@ *@$33B%B@BZC6@THmodUU>?rp>?DZ>???   % T`]qAA]LTmod % F|EMF++@ *@$33B%B@BZC6@H<NT*4>]?*>]???   % TX`-n;AA`9LPNT% F@4EMF++@ @$L MCZC43KB$B$$==% % LdF -!??% % $$AAFEMF+@<0Ne>@H<L MCXCCXCCZCL MCZCL MCXC@$$==_888% % V0I ss s% % $$AA( FEMF+*@$33B%BL MCZC@0$9=ARIAL6@H<SR0>L>r|>L>??   % TXAALPSR% FEMF++@ *@$33B%BL MCZC6@\PPre->>j`>>'>>ʢ>>??   % TdAA LTPre-% FEMF++@ *@$33B%BL MCZC6@THmodUU>?rp>?DZ>???   % T`AALTmod % F|EMF++@ *@$33B%BL MCZC6@H<NT*4>]?*>]???   % TX-;AA9LPNT% F@4EMF++@ @$(DZC03KB$B( $$=='% % LdF1* -!??% % $$AAFEMF+@<0Ne>@H<(DXCt5DXCt5DZC(DZC(DXC@$$==_888% % V0I1*s^-s^- 1* 1*s% % $$AA( FEMF+*@$33B%B(DZC@0$9=ARIAL6@h\</Ps>=?9->?jM>?>?>???   % Tl AALX</Ps>% F@4EMF++@ @$YCZC83KB$B$$==% % LdFQ -!??% % $$AAFEMF+@<0Ne>@H<YCXCCXCCZCYCZCYCXC@$$==_888% % V0IQs~s~ Q Qs% % $$AA( FEMF+*@$33B%BYCZC@0$9=ARIAL6@\P<Ps>UU=?r=>?>?1>???   % Td AALT<Ps>% F@4EMF++@ @$YCCZC83KB$B( $$=='% % LdFI -!??% % $$AAFEMF+@<0Ne>@H<YCCXCDXCDZCYCCZCYCCXC@$$==_888% % V0IIsv!sv! I Is% % $$AA( FEMF+*@$33B%BYCCZC@0$9=ARIAL6@H<SR0>L>r|>L>??   % TXAALPSR% FEMF++@ *@$33B%BYCCZC6@h\Post-q=>rG>>1>>>>q>>??   % Tl AA LXPost-% FEMF++@ *@$33B%BYCCZC6@THmodUU>?rp>?DZ>???   % T`AALTmod % F|EMF++@ *@$33B%BYCCZC6@H<NT*4>]?*>]???   % TX-;AA9LPNT% F@4EMF++@ @$DZC03KB$B$$==% % LddF9& -!??% % $$AAFEMF+@<0Ne>@H<DXC%DXC%DZCDZCDXC@$$==_888% % V0bI9&sf)sf) 9& 9&s% % $$AA( FEMF+*@$33B%BDZC@0$9=ARIAL6@H<SR0>L>r|>L>??   % TXuAAuLPSR% FEMF++@ *@$33B%BDZC6@h\Post-q=>rG>>1>>>>q>>??   % TlpAAp LXPost-% FEMF++@ *@$33B%BDZC6@THmodUU>?rp>?DZ>???   % T`rAArLTmod % F|EMF++@ *@$33B%BDZC6@H<NT*4>]?*>]???   % TXu-;AAu9LPNT% FEMF++@ *@$33B%BL CZC@0$۫>ARIAL6@TH...=?UP>?U>???   ( RpArialMonotype:Arial Regular:Vek%` 6` ` C` `N`Lw&w$:fw  w`whwpw 3`Bw w !  dv% T`#AALT... % FEMF++@ *@$33B%B DZC@0$۫>ARIAL6@TH...=?UP>?U>???   % T`/F#AA/LT... % FEMF++@ @<0Ne>@CZCCķECCC3CMC%CClC#CCOhCű CQC Cr:C! C#CC C%CLB Co BU.CB)?Cj?B'SC@$$==_888% % WX.t$ \ D= Q c    2 c Q  p 3 % % $$AA( F\PEMF+@<0f8B_OC BZC|AMB܋UCf8B_OC@( $$=='%  % V,(4  5Y  %  % $$AAFEMF+@<0Ne>@ CZC\CxJCACv{=CiژC 3CC]p+CXCF&C/qCӫ$C [C%CZ,FC-)C0CQ/C(Cm8CEC,EC(KBTC@$$==_888% % WPot $  t 1  lk K Y c [R A % % $$AA( F\PEMF+@<0=BPCf BZC{BɿVC=BPC@$$==%  % V,gt y 4l  %  % $$AAFEMF+@<0Ne>@th CZCױCNPC CuHCNCCC# Cw@C/ĔC?C;CbACG:C]EC}CKC`oC^TC@$$==_888% % WDt $ ; S j9    V  F % % $$AA( F\PEMF+@<0pmCYPCfCZC`rC7WCpmCYPC@$$==%  % V, i &t  %  % $$AAFEMF+@<0Ne>@`TCZCTC]TC-C̜PCêClNCgCCUC@$$==_888% % W<.t$ N  Y  '  Q % % $$AA( F\PEMF+@<0CPCYCZC CXCCPC@$$==%  % V,&3 a "  %  % $$AAFEMF+@<0Ne>@CZC*CķECa!CC3CC%C%ClCCC>*Cű C;D CV D! CrXDCD%CD Cś!DU.C\'D)?C-D'SC@$$==_888% % WXq$ \ = uQ  U 6!"$%c & g( ) H+3 % % $$AA( F\PEMF+@<0-D_OCF/DZCH,D܋UC-D_OC@$$==%  % V,f+ + +Y f+ %  % $$AAFEMF+@<0Ne>@ CZCCxJCCv{=C}_C 3CV C]p+C/CF&CCӫ$Cp1D%CܑD-)CI DQ/CRDm8C!D,ECDTC@$$==_888% % WPqv $ |  ,1  k 5K Y ! =# $ %R E'A % % $$AA( F\PEMF+@<0PuDPCF;DZC{MDɿVCPuDPC@$$==%  % V,q}^' ' 'l ^' %  % $$AAFEMF+@<0Ne>@th CZCaCNPCۦCuHCCCC0Cw@CuC?CCbACC]ECDCKCC^TC@$$==_888% % WDq $   9    V i RF % % $$AA( F\PEMF+@<0.;CYPCCZC C7WC.;CYPC@$$==%  % V,h  "t h %  % $$AAFEMF+@<0Ne>@`TCZCVC]TCC̜PCvClNC>CP:qAH̏މO&NW`3 $O贁Nk{L:"Qo_~4@Q_EhvPğ,3ꞒYǯ,x!1//gׯ>X3 U[mFԿٟRcXj|Ϡϲ 0BTfxߊߜ9K]oJUgyYq*<N`r!3E8J\ng! ////A/S/e/w////////??+?=?O?a?s?????sq???O"O4OFOXOjO|OOOOUbOnasfeO __-_?_@Q_c_u_______oo$o6oHoZolo~ooȡ I$ooo!3EWi{);M_qfjˏݏ%7I['>~Ɵ؟ 2DV¯ԯ .@+[mǿٿ!3fNrτϖϨϺπ&?P?b?t????8YrB;~^` OO?/1OW&POOaOsOOOOLOOO__%_7_I_[_m___?&G_E249_0,IoW&i0oBoTofo%I…o uooooo*<N`r_V*OJO.O@O1CUgyя+=Oas#/*UR8/\/ ,///7yӯ -?QcuϿS4$!laqYk(մҪћw3O贁Nk߬)x\%2qpì?o13? AKL"(΁iALQj(/ĕ+I:޿03)0k{+Ȉ+: !;Mz!3EWi//ASew-???=`5` S odX+LetterO_b PRIV|'rp38e4m/!/3/E/W/i/{//////// ??/?A=} DisplayUFDfP h> /T6DUmA@ ?ۿI?Y{O1@3EEbOeZ Hu P(:{W_qTFour-sided rectangle shape that can be customizby dragging any selion handle.mb?贁No?k?4 HD" # "=h-(>T;6UA?u?Q6 u`m u BAD00XA0(2s@sJLsT>5 LO@I5 I`?Copyright (c) 2001 Microsoft Corporation. All "s reserved.`Vis_Sba.chm!#26677Ad9 l>0>Udd !!T 6gbJC2 6D gs?2r4 ?`b"?U1f%a?z?%91U1|;6=!=5E2?;rA QBWCSq,^(;5/V:|L?C%=A,K?EFI=*3AONOOKAf%[6BWWV_Gr_U|\_>ZO_ Inha=J0Hgda*zTF!PF#f",s&exG#nRectangle,Four,sided,customizpdragging,sel{p* ,han}dpBasicpgnal,flow,systems,deq,control,ennpeerpPqf%QCT$@lrH'- !OyaGE-szF,# JB `aGao@+kbo3a\ oBP)UFDfP h>$/T 6DUA@ m?I? 3mU+!(:O@Wii +? sU !& )/?:eEZ1.D) (HoO/Ta/Q!2! ` Connector"`2 :. \?o  _# ?..:? ?j n`Uyek_k|f`eConnector that automatically routes betweene shapit cs, using a curved diagonal line.b]ٿXRx??L&d2?Pq߿X){Dv? i?% H DH D # =h8>T  PYY9 #yAU@?;?@} P!3|@u `u`bwu  - :u`h?mu`b"Y@0S#  AS.A$ A[mY,'>މU@dx-E?@d2L?@?Ȣ?@m6?f?!n'%"x//'/%/B*E4$!g tNzsS K2 Q< 83&7 0Q46 & " "2R z6 LS --1?2rq]I3@(EM1E??r\BbCg*+0`A*@baC SBuUAbWuha4~#`muY`}@ }bPu`#28& "S=#WQ@1@37 `Vis_Sba.chm!#26674I`?Copyright (c) 2001 Microsoft Corporation. All Rs reserv?ed.1TT QyP0l>(>U@ m5 1(X=$9"8^=$L??<bE]fv8 agj92]nha| .u9 TT@1h Ac`Change Arrowhead..P\Bb)Seq4%7@1WQ}[]Jf WqEx ,E`@`_Set As StraRLine-'szF< )#C&B t'm~a _@+k KaL(kV:t4:m@t@ @ dK}A-t7"AU)*+t4:m@t@ @ <C- 7AB@IR@|LRH<( H<( BE| RE R(vlSm+B!/?D9CRPD,aAM3AM3PPTPPT1h/F~ A/,F A/DF A/\F AV7V @ ?g "*^p? N(T*GK! LCUP ( @ C?FuDTe]y ahZ$T qUI:m@It@ @A?I?.CUgB^ a*  ZV^ +W /-%^U*-?-(rI!u` h"""ua)#R$%B6L>^>B8sU16t?$"o%<}'Q#( A41 A \"` Connector`+-B 3Je?d?UGHFktAA@|A{!* @$2nU.T6?A7!Qt!A {a7?Rt1R;;UWyUTWvbU' `P|(,5A[ 8# ,h @5!@?!$P&arentSl0!RUQ atA 0^%a|(g!/ 8kV u0 stEoRzczq^,!Wg!uoawT6/Wqq@{VWx)#5/)_htE"}gqD$TTU}'^TiAiCu;M_ߕߕCO$7?_[mO  !\Sd_L^\1a8bW8eoo_:@L /pT)u!ky#l/{ %/////// ?P+e'?9?K?]? o??:0???>OO*O ߭UOgOyO_OO]_OOO _-_aQ_ӯu_____3o)ooMo_oqome*ϿoǏҿ=Oasm -#W#5GYk}я:A?Qc>004.[ bbbz 8ISR Pre-mod  NTI7Imǯ믍!3W{ÿտ φjeVϴ^ϰ9 .@ as25߀&81@Oas =rF}fFB5GYk}1CU_y_?o_ [/-/QouA/O< /koO@U/g/o///////0 Y?*?P5M?m?5????? ?1sOO)O;OMO_OqO@jFXOObbAOIs _u3QCX]_o__o__eo_:_o#o5o.Yo}oooooo 1Cgy$jHZl~, ݏ) _7C$uH@Zl~Ο(T,?4rODVONQO ί((ֿL^pʿc6TZx~:Rψ}uo/B /L^ߔ߸߲`?9CK D1d O$ 2DV4h|AsbVaR. n"bbb/zR %SR Post-mod NT4FX|<Uo ?0oTfxo///,/>/P/b*b_%'/oO//?+?=?O?a?s??I@5Y?x11 ?V璑?ğO@1OCOUOgOyOtOOOOOOO+PRC_-_?_&z________ o@.o@oRodovoo=Ͼoo*<NrďaϺ%S[7я+=l0* ]2J|N  0fwSew=ӯ0審28pc[7\#`0O贁Nk%,p y...ȿ"ϹFXj|u7/ς ^0BTfxߊ?I#5GK?k/CgO$IJiAYC"%5V!mj'2UR՝BON`o$3K+=Oas?cuU0{Gz)R!2DVhz? ?/@/R/d/v//////5/?*?2h?(ʗ??C8?O? BOpd4YEX]OoOOAҗJζζOOOO__&Ud2_USmVd_v_____~潥__5ٗoKoϵ5ȃoooooo,7I[mu Ɓc #h|YnYkŏ /ASewᵟ$ҟ,>Pb* D>؞z:ůׯK7yp8=&nD:`D?P81,>PbtD7?P?b3~To??4ў?Koq????O,O>DMO_O#сO#Om vO__%_7_I__m_//__}_t!o3mTofoooj?o\??oGY1/ mx? A#_ 0`BTfx!HA E.@Soefptefo@on$ADvvRD'9KpVh PIq߀UuA9??oߙx/B/%/Zl[/m///////-v??)?YG??}???@ OOKOoOOOOOOOAO_Y6_H_Z_l_~___*a[_oo4oooojoߎh=CֻoooDD20BTfxJh sv"4F3Wi-U /ASwџ+-n.;ؙޯϯ(/;v]o/ɿۿ/+TI[mϑϣϵTJ^. Z7I߽Ztߘ߷Op1EVO_ - rHzVzVhBfxu1΋I@s,>Pb>iEiQQ/c/+oJE4h/ASew+=Oa=4f-Ǐ7H/(/^;/q////R///?#?5?G?e1pqFe<?pqe??3F15| ItOJE@@ƣq EWƣ?Qytwq@tnƣEq|wPQuTc^⥂!/FFR#bx{~Y䅥\J~ ~l>Q,DϪ=Οـ|u+YJE@ڟyq\׿䅂+Jۨ~S/>QSN`䂨ah~Id?Ư pqGv1#8 5}; I>Pѿ:m@t@ P XcJTJprv}pru{x20v1rpqsQsuvRsnss QFs4yx#1s\As"pvasQsf VPmp U   4U4HH\U\ppU5e4pqcXpqBHD(=Q v6$QQ;Upq3u4Q7q47䅓4>Q7J4n74#7F47U43U4K17e4O17ve4S17Q8va78Q7c187g18c17y8g1764y7Q86787{18{17LՓ4ў7e4a3"q"?# 96an1#!Pa&"aC~ajaK1$waO1%ЄaS1&БaA#'Оa!++Ƭ)иa0Q2ag13ay#4ao21aF-a{1/qQAApqg1S19UnU U!"#$U%&'(U)*+,U-./0U1234t4:m@t@ @ >[C-K/?A@J?MRH<( EJ? R\K!I>@?Il?.PDK?.PU1( UO"D&aU=QJf )h"Ty+U-x_Ʌ&!Q- H*9(TYgTEQ/,GuideTheDocPage-1Gesture FormatRectangleVisio 90ConnectorVisio 00Visio 01Visio 02Visio 03Visio 10Visio 11Visio 12Visio 13Visio 20Visio 21Visio 22Visio 23Visio 50Visio 51Visio 52Visio 53Visio 70Visio 80BasicBasic ShadowvisKeywordsvisVersionRectangle.2Rectangle.3Rectangle.4Rectangle.5Rectangle.6Rectangle.7Rectangle.8Rectangle.9Rectangle.10Rectangle.11Rectangle.12Line-curve connectorAntiScaleScaleRectangle.22Line-curve connector.18Line-curve connector.19Line-curve connector.20Line-curve connector.21Rectangle.17Line-curve connector.14Line-curve connector.15Line-curve connector.16b553@ E3A E3A E3AG31AE3?AA3CA E3PAE3^A E3,kA E3DxA E3\A E3tA E3A E3A E3A E3A E3A E3A E3A E34A E3LB E3dB E3|!B E3.B E3;B E3HB E3RBE3cBE3sBE3BE34BE3LBE3dBE3|BE3BE3BE3BE3CE3CE3 $CE3$5CG3DTCE3\bC E3llCE3}C"G3C"G3C"G3C"G3DE3D"G3<8D"G3\ZD"G  !"#$%&'()*+,-./01234U^U U UUt4:m@t@ @ *IKC-u 7A%t4 o_I A-_7AJ@{ILR@I6RH<( H<( JE,3K RE$@K R{  g"4FX (h(NT1@(߹A7N1y [La:]B_ P  bw&5N!,co >;F}S"o0FuaR$|EbxP.ġx|NtF+NR)N+'$G?h14=@===U2 |DT| J*P# MK6%?!`<̇wL C|k @                        ! " # $ % ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < = > ? @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z [ \ ] ^ _ ` a b c d e f g h i j k l m n o p q r s t u v w x y z { | ~  Oh+'0@HXdp|ringgerG_ EMFl@VISIODrawingL _` ??d((`hhh@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@PPP ߿߿ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@  ```````````````````````````````````````&&&LLLLLLLLLLLLL&@09@:`LLLLLLLLLLLLCp 000``````````````````````````````````````````==3MMffffffffffffffffffffffffZZTTT````````````````````````````````````<<<]]]09@:`LLLLLLLLLLLL:`09@0LLLLLLLLLLLLL ```````````````````````````````````````&&& 444KKKL09@s ``````==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ``````==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKssL09@sss ``````==3&333333333333-xxx]]]09@ssss09@ >`|s  444KKK 444KKK_ L09@s ``````==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@ss ``````==3&333333333333-xxx]]]09@sss09@ >`&@L  444KKK 444KKKswL09@sss ``````==3&333333333333-xxx]]]09@ssss09@ >`ss  444KKK 444KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444````````````KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKK|LLVL09@sLLL ```==3&333333333333-xxx]]]09@sLLLs09@ >`LLL  444KKK 444KKKLQQQ%%%QQQ09@ss ``````==3&333333333333-xxx]]]09@ss09@QQQ%%% >`&@L  444KKK 444KKK_ L09@s ```xxx````````````==3&33߀333333-lll`````````xxx]]]09@ss09@ >`  444`````````xxxKKK 444KKKL09@s ``````==3&33333333-HHHxxx]]]09@ss09@ >`  444KKK 444KKK|LL_L09@sLLL ```000```==3&33333333-xxx]]]09@s_LLLs09@ >`|LLL  444HHH000KKK 444KKKL09@s ``````==3&333333333333-xxx]]]09@s&@s09@ >`  444KKK 444000000000000KKKn&@&@CpL09@s&@&@&@ ``````==3&333333333333-xxx]]]09@sCp&@&@&@s09@ >`n&@&@&@  444KKK 444KKKL09@s ``````==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444000000000000KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ```==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKK&@&@L09@s&@&@n ```==3&333333333333-xxx]]]09@s&@&@|s09@ >`5X&@Q  444KKK 444KKKLL09@s ``````==3&333333333333-xxx]]]09@sss09@ >`  444KKK 444KKKLLL09@sLL| ``````==3&333333333333-xxx]]]09@sLLs09@ >`VLi  444KKK 444KKKL09@s ``````==3&333333333333-xxx]]]09@ss09@ >`  444KKK 444KKKL09@s ``````==3&333333333333-xxx]]]09@ss09@ >`  444KKK ```````````````````````````````````````&&&LLLLLLLLLLLLL&@09@:`LLLLLLLLLLLLCp 000``````````````````````````````````````````==3MMffffffffff ffffffffffZZTTT````````````````````````````````````<<<]]]09@:`LLLLLLLLLLLL:`09@0LLLLLLLLLLLLL ```````````````````````````````````````&&& kkk444ggg{{{,,,???gggOOO(((KKKoooccc 444[[[```WWWEEEggg<<< KKKsssSSS000<<<oooGGG$$$sssooo444ccc pppsssPPP"""```###&&&:::xxx (((;;;PPP NNN'''(((``` ppp@@@ǏXXX___ HHH翿XXX```000PPP000ttt000KKK @@@```888```hhh@@@@@@(((888eee---  kkk@@@@@@@@@XXX000PPPXXXHHH888 HHH(((xxxppp ߏ666 (((333+++ $$$000PPP(((hhhXXX 000(((ppp000ppp((( WWW%%%PPP@@@XXXPPP"""OOO***pppﷷxxx@@@XXX000xxx@@@ @@@PPPHHHϏXXX ```hhh@@@@@@ 000QQQ888XXXhhhKKKZZZ$$$SSS@@@@@@@@@```hhh000HHHhhh888000 pppXXX000ߟ߷HHH888@@@888MMM 000翿 HHH 888```(((HHHWWW```hhhhhh@@@ 000888((( ```pppTTTwww(((PPPxxxhhh888(((((( XXXXXXhhh`````` (((```  PPPhhh```888 xxx888(((秧hhh000(((888hhh``` @@@ ```888(((  hhh@@@ח``` XXX```888```PPP@@@ߟhhh(((PPPׯHHH ```(((ppp xxx 翿XXX000888```PPP(((hhh(((```(((ppp篯ppp888@@@ppp@@@(((PPP000XXX ```000ǟxxx@@@@@@@@@@@@@@@@@@@@@@@@PPP000xxxZZT::4(((XXX@@@@@@@@@@@@@@@@@@@@@@@@hhh@@@HHH PPP@@@HHHHHH880``pphhh888XXX000 888PPPxxx @@@880``ppPPPhhhhhh @@@```珏((((((880``pp888 xxxhhh 000 PPPXXX@@@ppp880``pp߇(((000```@@@ ```@@@ǧXXX``````880``pppppHHHXXXPPP hhh 000@@@```￿```@@@888XXX```880``pp```@@@@@@PPPϿppp@@@@@@``` ϿPPP@@@888@@@888@@@XXX```880``pphhh@@@000@@@ 000@@@@@@ 翿```880``ppϿ ```880``pp ```880``pp ```880``pp 880``pp 880``pp 880``߀pp 880``߿pp 880``ppp ```880``pp ```880``pp ```880``pp ```880``pp ```880``pp ```880``pp ```880``pp 880``pp 880``pp 880``pp 880``pp 880``pp ```ZZT::4 ``` ``` ```@@@@@@@@@@@@@@@ `````` ```@@@@@@@@@@@@@@@ ``` @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ ǿ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@pppPage-1 RectangleLine-curve connector8_VPID_PREVIEWS_VPID_ALTERNATENAMES_PID_LINKBASE A Oh+'00 x      (-Amalgam: A machine-learned generation mo3yXXF/=ZvDʱ}vt``[ 0Uk-YؑlЀ]ȳS)(N%aWAm~ =9BA<0Lh֓je㫥[׋ۖx^AWhB{TUg@/qɺZ[wba[9V Y2 Sqd,a"7P\XԙK\tl9CXn˗w"!WcXwK`2йx g;l^m=Cʂ8 Vx]p{ RYloAζʃ'(o] aϮ^l2ؼ;I`f]ǰ+7d ātR-Ypa9}i΂k7 ;`q]Xc?խ'^syrj¾ik@}ln+|!Pkf朰gҰ㲉'sbٺTtMR)ξ7ԃzlWcQ)vOփr:kM<_%KU"k ܬG.qۘEUoHz,SLgHª\!Z f`Պ<,؈-փBD=(cbʯ鬆Jj&GKvVnc+WgV]ƕa;@ڲA}Vc)`q&.Z J V׃i} mYj\qҧ) [kty"x{Qz z\lGbMld3ӖFש݃=>m Xa `MJilZmm} ldZ_֧kש}٬dӌdO[ ˸f3/ +S=>mX?Td˺קe;6%[Wm?=Cn9ϦtIENDB`}DyK _Ref536603733 Dd x  C 4AafterorderingC"b i2 VVZr Hn i2 VVZrPNG  IHDRIsRGBPLTEg&T aIDATx] |`=q+|9ZKoNkW' <^ex|o\xP ׏#Z/ox9='羃z9YZ6J ;Zk?:1oZ}r  T) FyQ}441WTr7/]GQy?fD[n+- \?>z@v60ZWY`ǶTb/~_3aUʚEװDŁG \yjx7j/p̫ao.0u͋HҰWe $p]` IݒGF<<ݒT--w:~Ez1劅k,X/]}*yeKW[P >X8 =ӈ=V7[]%G5 2eԐ =?\d2$}%UV4na =c(hpȑVHl˜"E݀&yt0ݕ.fx2j5q!τ $ý |G]UGU5ҩS>gW]ŬJ3ƴ#kP?ՈxSjXY`E洤bӺxȜ˝Q_DU| - 51U׺,@xO38``VN bGnT?|=Ƨo#?2+8۽ʟMz}+hNyXU 4leȇPZAgv)x61vƹaF/Ecmn5l O# GM*rB )`Va7k7dG$9U 3`4 jïت~A6 ,g]5 V#5ֵ5 WT3 zg ݆+`F`x61'$5Z/Z=8~0.ZB} Mk0P~ȿ+2^q+њv1d٣1_F^--Pa$rƼr!&h}1S86VsWBNO5Ŀ*!R rxHhGPh6,@&^^mv|w,73zSؚQMٍ6/^Zځ8cz?[]t>7|VùS^ W71Ԃ_#6m-M Ae {/i15`xtĽ#2Ӧ/x!J&HxM  8 '7 2mpl_c*^q K=ۗ ˫PFޛIkp>1$Q=͒,+ڛIOKcjo+q^} `* =)Ia׭+4YѴw :P+pj+Û 1`p4b2匽t>{8I^miL4d/\VMl\e"xx:ڤPik-7g%~p1+, |7װonӅߦ&}+s;_kif1|03F?qsx#F]FYߗbu4*_WB/֖>Xe7j=IEu-l V\d#d;s7bG3bɖ`#j84'v=4> 6M:mQi9Q<=&_|H&+ss > ]D{l<e۽XysAes=;9#{uh8M~ϛ+d.ӣfGMO]560 E!wJIH/)fȎ$le$.` &4KJdݭ;^M=:)E0p{blAƘCK݈ȍ#73_ơo|-xnsn CŸ!|RQ@mJ݀Fitӹn^8F`N1/ڢoاj} S:So/ eId¸eJ7dr4R‹p ؜[Õf1Vg uaF9% srJ;v8t1b>r+d}h_hz2숶 ;ڈ;C}lXNcTJTb{NNHC=>$A\μ,x2<:0|-x5E*ËE #X2Iu@B//n/:)r7wjR jP&6gQO-fvo <5Z˶4GO?>n\}S6NkEs!0 ,Wi K5K[Ihid'SQffϸ5^ S^ruc\&3g&3? 2y]Of@V : 0Aq_9b'=H0m!S_N H>a1G62\ko _.gn>\Q/44:O@]sn5{|.\skϽ`ZUZ|n-,.Jyu=|9 >X'\/fwcWW_\rTD sNj7qX:Q_~9*p_ 5{\~ҽ/{RuMk_lm[/lEP\b {%{} _u&rֈW=ûX/ 5{\~jףּUef6Pt(X+@:j=ooy6{녊`kO3pgHnD49 $ /E<1ү0?$Nu.jZ77){WDdfCg\PAHW1~-OZ%LJM|)N8B5g]/o;=igZ8C5Z[TSWOq\~ʌ3* X=zΓ~ƚX2ZJ|ty;1}v諸nG㭤ۖ;@,Q=j8ZoGc֗z'psv1=ۇŞ1SW/8`wl/+q|g<~KV88p۸U{$=>S\zl sw&̌EI^*b{odHg=Lқn9z!VKvJ2KNΎ {ԸHk4u۴4LqpiO}ͷC6ђ13߇sF\vӆ$y;c 15L0$% y7vnrֵ\r~|KؠB4Im%(+egiF,f-y缆zgN|$<{JxW)o'w ?u .KƇW+M>>!xF;m83xrqڎ0y+#1۩qۢ/2Fܮ|i)[;EV}qOzNأI{4]|,-̍e Dwa{4ae5 jyךCpӱJ<VW>qZ\YY߮0om4~7w.O֣F>ݿ:e:?]/,C~F_nkOYcN;{$wnNo,Fצgn{0 =%a/aB$(;YLx5fz|ܛxqqtow%mHӯLLŶAixݬ8n}>`٤֢ HA~봆iE7V2|3ԗ꥟\2:],8Y*ݎ}<ˡB.tq&ޮ %(˿t/4(8B!|Ջ4þ5B>U5B>P0@M}|9aF~h\AVzcu|2}3?ߩ=pfCJG9͓!S/|=~I7oW7Bܟ3GF3B>;NrKli}}Mkl ʷb_st7_4tW4 8Ec'Xa4Ϗ׮p0B??ƺ!|~k/`xnuH'Jgڣ}& [{:YmlUsuBꇇu~>baI5B;抲xTa ҁ0d-AVTEc#i໏i{4kQW8Ec5~Gc/XSXu:eX_$1__{諽5}~ӭЗ۩~䰫4_;]?Xw_dht̟GEoX?.}WXXcaXY#MJ{z}>MrZΏ1ۏQy}gߙ.<'[,.i󎓢QMO5mFXaDsbP^C}DyK _Ref536603861}DyK _Ref536603888yDyK  _Ref8111122}DyK _Ref5315045671Table' wSummaryInformation(2t`DocumentSummaryInformation8 CompObj~j=N@N ~|Normal $a$CJ_HaJmH nHsH tHt@t  Heading 1*$ & F h8<@&5CJ KH OJQJ\^JaJ d@d s Heading 2$ & F<@& 56CJOJQJ\]^JaJr@r  Heading 3+$ & F h8p<@&5CJOJQJ\^JaJN@N +4 Heading 4$<@&5>*CJ\aJT@T s Heading 5 & F<@&56CJ\]aJN@N s Heading 6 & F<@&5CJ\aJ@@@ s Heading 7 & F<@&F@F s Heading 8 & F<@&6]T @T s Heading 9 & F<@&CJOJQJ^JaJDA@D Default Paragraph FontRi@R  Table Normal4 l4a (k@(No ListV>@V sTitle$<@&a$5CJ KHOJQJ\^JaJ .O. [ ReferenceB'@B gComment ReferenceCJaJ<"< g Comment TextCJaJ@j!"@ gComment Subject5\HBH g Balloon TextCJOJQJ^JaJF"@F NCaption$xxa$5CJ\aJj@cj kiD Table Grid7:V0ZOZ i; example sentence$p`^p``m$6F^@F ]:M Normal (Web)dd[$\$PJZZ\bKSimon Corston-OliverCJOJQJ^JaJph>@> 7a Footnote TextCJaJ@&@@ aFootnote ReferenceH*O:| PseudocodeG > @`0Pp @ CJOJQJmHnHu^O^:|Pseudocode Char'OJPJQJ_HaJmHnHsH tHuTOT:|Style Pseudocode + Bold5\VOV:|Style Pseudocode + Bold Char5\B@B (TOC 1 $xa$5CJ\aJJ@J (TOC 2!$x^a$6CJ]aJ@@@ (TOC 3"$^a$CJaJ@@ (TOC 4#$^a$CJaJ4 @B4 (Footer $ !.)@Q. ( Page Number4@b4 (Header & !@@ sTOC 5'$^a$CJaJ@@ sTOC 6($^a$CJaJ@@ sTOC 7)$^a$CJaJ@@ sTOC 8*$^a$CJaJ@@ sTOC 9+$^a$CJaJ6U@6 ( Hyperlink >*B*phNON nIndentedQuotation-hh]h^h,O, {Citation.LOL SingleSpacedExample /$$XOX P! References0x^`CJPJaJtHTOT 2GText Indent Char1 `  CJPJtH^O!^ 1GText Indent Char CharCJPJ_HaJmH sH tH4O4 5GText3 CJPJtHLOQBL 6G Caption Text 45CJaJtHFOQF 3G Text CharCJPJ_HaJmH sH tHZOaZ 4GCaption Text CharCJPJ\_HaJmH sH tHTOqT GFootnote Text CharPJ_HmH nHsH tHRYR z Document Map8-D M OJQJ^J@O@ &CDoublyIndented 9^<O< &C MinorHeading:$6FOF UnnumberedHeading ; & F$L@$ Date<Jn۔  hO} BSVI      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEF}}}}}                BSSSSSSV  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGH hO                hOohOp.pcppppI z z "z z z z z z z z$z z$ z$ z$ z$ z z$z$z$z$z$z$z$z$z"zzzzzzzzz z!z"z#z$z%z&z'z(z)z*z+z,z-z.z/z0z1z2z3z4z5z6z7z8z9z:z;z<z=z>z?z@zAzBzCzDzEzFz?&2 ?KW\bWkAu}ߔW3C%o2-0"*V4;BMI P5W_ hxpzA;dټ W.h  F*3<yFhO\it  9  B 'Yc7J9%W !"#t$%&'(!)+*p+,-#./<01L2"3 45678P9 :;W</='>?@A&BZCDEFGH -.d't[E!&)).]0267|89t:?BDyHPLM7PQQJT#VFVeXXXY>YYYlYYYY%[=[Q[e[[[___````7`8`=`B`H`M`V`^`_```a`u`v`{`````````````````````` a aaaa!a*a2a@aJaKaLajakapauaza~aaaaaaaaaabbbbb$b)b.b/b0b1bGbHbIbObTbbbobpbqbrbbbbbbbbbbbbbbbbbbbbccc$c%c&cCcDcUcVcWc[cacfckcwccccccc dgghiilmnNqqr_rrs=stssssuvvv wWwlwwww4xdxz||||}}(}8}9}?}F}L}S}T}Z}`}f}l}m}t}z}}}}} KUHrAC*Xϋ )L-xǎȌp Q]uvޔ6Öf8ߞA[hS{æ3Gc ڸY1)}Iy24 gR $   6 //////1020_0`000000011E1F1j1k111111122L2M2222222334353g3h333333399lFmFLMMMeVfV]]6aAaDdSdddd eeee'f9f]f h0hhhhhiiijMuNuWX 78MNiO 0 00 0 0 0 0 0 0 0 0 0 0 0 0 @0  @`0 @;0@0w@0w @ 0@0@0@0@0@0 @ 0@0)@0)@0)@0)@0)@-0)@0)@0)@0)@0)@0)@0)@0)@0)@0)@0) @ 0@0Q@0QA 0QQ@0cV@0cV@0cV@0cV@ 0cV@ 0cVA 0~YcVA 0~YcVA 0~YcV@ 0cV@ 0cV@ 0cVA 0e[cVA 0e[cVA 0e[cVA 0e[cV@0cV@0cV @0cV @0cV @0cV @0cV @0cV @0cV @0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @/0cV @0cV @0cV @0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @0cV @0cV @0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @0cV @0cV @0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @0cV @0cV @0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @0cV @0cV @0cV @0cV @0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @/0cV @0cV @0cV @0cV @/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @/0cV @0cV @0cV @0cV @0cV @0cV @/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @/0cV @0cV @0cV @0cV @0cV @0cV @/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @0cV @0cV @0cV @0cV @0cV @/0cV @/0cV@/0cV @/0cV@/0cV @/0cV@/0cV @/0cV @0cV @0cV @0cV @0cVA 0QQ@07d@07d@07d@07dA 0QQ@0iA 0QQ@0/m@0/m@0/m@0/m@0/m@0/m@0/m@0/m@0/m@0/m@0/m@0/m@0/mA 0QQ@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv@0Bv @0Bv@0Bv @0Bv@0Bv @0Bv@0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv @0Bv@0Bv @ 0A 0OO@0a@0a@0a@0a@0a@0aA 0OO@0@0@0@0@@ 0@0@@ 0@0@0 @ 0A 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 0P 00(@0J @0J@0J@0J@0J@0J*B 0JJ@0-@0-*B 0JJ@0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0@ 0 @0*B 0JJ@0y*B 0JJ@0K@0K@0K*B 0JJ@0@0*B 0JJ@0@0*B 0JJ@0*B 0JJ@0 @0  @ 0A 0@0@@0@@0@@0@@ 0@@ 0@@ 0@@0@@0@@0@A 0*B 0AA@0e@0e*B 0AA@0U*B 0AA@0*B 0AA@0*B 0AA@0^*B 0AA@0*B 0AA@0G*B 0AA@ 0S@0S*B 0AA@0"*B 0 AA@0&@0&@0&@0&*B 0 AA@0O-*B 0 AA@0q0*B 0 AA@0N3@0N3A 0@06 @ 0 _y00Xo.@0 _y00o.@0.`_y00o.@0_y00p.2!_y008`SD@0_y0 08i@0_y0 08i@0_y008i@0_y008i@0_y008i@0_y008i@0_y008i@0_y008i@0_y008 i@0_y008Xi@0_y008i@0_y0 08Ti@0_y0"08i@0_y0$08Ĕi@0_y0&08i@0_y0(08x_Y@0_y0*08_Y@0_y0,08_Y@0_y0.08 `Y@0_y0008X`Y@0_y082I{009/:0 _y0:24J{009/:0_y0:24J{009/:0 _y0>2J{009/:0_y0<2lJ{009/:0 z082lI`{009/:0{009/:0{00:/:0{00;/:0{00</:0{00=/:0{00>/:0{00?/:0{00@/:0{00A/:0{00B/:0{00C/:0{00D/:0{00E/:0{00F/:0{00I/:0{00K/:0{00K/:0{00M/:0{00M/:0{00O/:0{00P/:0{00Q/:0_y0Y2M@0zz_y0[2M@0zz_y0a2xN@0zz_y0]2N@0zz_y0_2@N@0zz{00R/:0@0zz_y0i2XO@0_y0c2N@0 _y0e2N@0 _y0g2 O@0 0R4iO@00 -.dm)>G'~BJV' H  ^ + e 0`*?^' d0RYbw't[E!&)).]0267|89t:?BDyHPLM7PQQJT#VFVeXXXY>YYYlYYYY%[=[Q[e[[[___````7`8`=`B`H`M`V`^`_```a`u`v`{`````````````````````` a aaaa!a*a2a@aJaKaLajakapauaza~aaaaaaaaaabbbbb$b)b.b/b0b1bGbHbIbObTbbbobpbqbrbbbbbbbbbbbbbbbbbbbbccc$c%c&cCcDcUcVcWc[cacfckcwccccccc dgghiilmnNqqr_rrs=stssssuvvv wWwlwwww4xdxz||||}}(}8}9}?}F}L}S}T}Z}`}f}l}m}t}z}}}}} KUHrAC*Xϋ )L-xǎȌp Q]uvޔߔ6Öf68ߞA[hS{æ3Gc ڸջWY_/1n%Ro)}Iy24} gR $   6   >R0ETez.;K_x4Gp!""^##$$&{''').)v)))%*C*U**S-v-////////////////00*0102080H0X0_0`0f0r00000000000000001111.1>1E1F1K1W1c1j1k1p1111111111111111 2222%252E2L2M2P2d2x222222222222222233 33-3435383L3`3g3h3n3~33333333333333V4g46&616666%7M7777h88s9w999999999999999::":):*:1:D:W:^:_:h:{:::::;f<z<<==9=x=== >><>>>Y?]?g?n?x?y?????????????@ @ @@%@9@@@A@F@Z@n@u@v@@@BB%BC C_CCCD)DJDOErEMFQF[FbFlFmFrFFFFFFFFFFFFFFFeGvGMImIxIJJJKHKiK{KKLLoL-M1M;MBMLMMMPM`MpMwMxM}MMMMM*N;NL[bckyѶ߶  !,:IPQWgw~Ʒڷ 4HOPUi}ø׸$+,1EY`akyƹǹι '(0BT[\dv̺ں  '5CJKSfxϻ 2DKLTbqxyѼؼټ -45<N`gho½ɽʽӽ#*+1?MTU\l|;߾ -<CDI]qxyʿؿ߿)7ELMUcqxy*/8y2"E%,-6DRYZ_o  "078=Oahin#*+0CV]^ft$29:AQahirdu$[&uSe&3@GHs/,;W2v&-78?Ukrs{189@Rdkl.2<CMNUk!");MTU2(;UE9h~!>~N}9Xn P[   =(zLX!   !##s$ %%&|'6((j)F*++`,,"-7.3//00B1123'4466^777889,:p: ;R<<=>z?@BB`CDoDDEyFGG9HHTIIJKKmMN*O+O?O@OIOJOKOLOMONOOOPOQOROSOTOUOVOWOXOYOZO[O\O]O^O_O`OaObOcOdOeOfOiO000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 !0 !0 !0 !0 !0 0 !0 !0 0 !0 "0 "0 "0 "0 "0 "0 "0 "0 !0 0 !0 !0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 !0 0 !0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 "0 !0 "0 "0 "0 0 0 0 0 0 0 `0;000 000000 00)0)0)0)0)-0)0)0)0)0)0)0)0)0)0)0) 00Q0Q 0QQ0#V0#V0#V0#V 0#V 0#V 0>Y#V 0>Y#V 0>Y#V 0#V 0#V 0#V 0%[#V 0%[#V 0%[#V 0%[#V0#V0#V0#V0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V 0QQ0c0c0c0c 0QQ0i 0QQ0l0l0l0l0l0l0l0l0l0l0l0l0l 0QQ0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u 0 0  000000 0  0H0H0H0H@ 0H0H@ 0H0H0H 0 0ϋϋ0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00000000000( 00000( 00 0 0 0 0 0 0 0 0 0 0 000( 0000( 000000( 00ջ0ջ0ջ 0ջ( 00000( 0000( 00000000 0ϋϋ0%0%0% 0 00000 0 0 000000 0( 0}}00( 0}}0( 0}}0( 0}}0( 0}}0g( 0}}0( 0}}0R( 0}}00( 0}}0( 0 }}0000( 0 }}0( 0 }}0 ( 0 }}00 00  00  0  0 0 0 0 :0 90 :0 0 :0 0 :0 90 :0 90 0 :0 0 :0 0 :0 0 :0 0 :0 0 :0 0 0  0  0  0  0 0 0 0 ( 0  80{'0'80{'! 0)! 0)! 0)! 0))! 0))80{'0C*0C*80{'0S-0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S- 0S-80{'0V4( 0  8060&6806 06 06 06 0680607078060h80h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h8 0h88060:( 0  80f<0z<80f< 0= 0= 0= 0= 0=80f<0 >0 >80f<0>0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0> 0>80f<0@( 0  80B0B80B 0C 0C 0C 0C80B0D0D80B0OE0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE 0OE80B0eG( 0  80MI0mI80MI 0J 0J 0J 0J80MI0iK0iK80MI0LL0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL 0LL80MI0*N( 0  80 ? @ A B D E F G H I J K L M N O P Q R S T U V W X Y Z [ \ ] ^ _ ` a b c d e f g h i j k l m n o p q r s t u v w x y z { | } ~  0 0 0 08000( 0  80s09 09 0080s0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0 : 0 : 0 : 0 80s080s00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 080s0 0  ( 0000000( 00(0(5 0(5 0(0(4 0(4 0(4 0(4 0(0(B 0(B 0(0(C 0(0(0(0(0(0(0(0(0(0(0(0(( 00X0X 00 00P0P0P0P30P30P30P30P40P 0 00000 0 0000 0 0;0000000000000000000000000000000000000000000000000000000000p0000&00$0$00$000000000000000000000000000:00. -.d't[E!&)).]0267|89t:?BDyHPLM7PQQJT#VFVeXXXY>YYYlYYYY%[=[Q[e[[[___````7`8`=`B`H`M`V`^`_```a`u`v`{`````````````````````` a aaaa!a*a2a@aJaKaLajakapauaza~aaaaaaaaaabbbbb$b)b.b/b0b1bGbHbIbObTbbbobpbqbrbbbbbbbbbbbbbbbbbbbbccc$c%c&cCcDcUcVcWc[cacfckcwccccccc dgghiilmnNqqr_rrs=stssssuvvv wWwlwwww4xdxz||||}}(}8}9}?}F}L}S}T}Z}`}f}l}m}t}z}}}}} KUHrAC*Xϋ )L-xǎȌp Q]uvޔߔ6Öf68ߞA[hS{æ3Gc ڸջWY_/1n%Ro)}Iy24} gR $   6   >R0ETez.;K_x4Gp!""^##$$&{''').)v)))%*C*U**S-v-////////////////00*0102080H0X0_0`0f0r00000000000000001111.1>1E1F1K1W1c1j1k1p1111111111111111 2222%252E2L2M2P2d2x222222222222222233 33-3435383L3`3g3h3n3~33333333333333V4g46&616666%7M7777h88s9w999999999999999::":):*:1:D:W:^:_:h:{:::::;f<z<<==9=x=== >><>>>Y?]?g?n?x?y?????????????@ @ @@%@9@@@A@F@Z@n@u@v@@@BB%BC C_CCCD)DJDOErEMFQF[FbFlFmFrFFFFFFFFFFFFFFFeGvGMImIxIJJJKHKiK{KKLLoL-M1M;MBMLMMMPM`MpMwMxM}MMMMM*N;NL[bckyѶ߶  !,:IPQWgw~Ʒڷ 4HOPUi}ø׸$+,1EY`akyƹǹι '(0BT[\dv̺ں  '5CJKSfxϻ 2DKLTbqxyѼؼټ -45<N`gho½ɽʽӽ#*+1?MTU\l|;߾ -<CDI]qxyʿؿ߿)7ELMUcqxy*/8y2"E%,-6DRYZ_o  "078=Oahin#*+0CV]^ft$29:AQahirdu$[&uSe&3@GHs/,;W2v&-78?Ukrs{189@Rdkl.2<CMNUk!");MTU2(;UE9h~!>~N}9Xn P[   =(zLX!   !##s$ %%&|'6((j)F*++`,,"-7.3//00B1123'4466^777889,:p: ;R<<=>z?@BB`CDoDDEyFGG9HHTIIJKKmMN*O+O?O@OIOJOKOLOMONOOO^O_O`OaObOcOdOeOiO000 0 0 0 0 0 0 0 0 0 0 0 0 0 `0;000 000000 00)0)0)0)0)-0)0)0)0)0)0)0)0)0)0)0) 00Q0Q 0QQ0#V0#V0#V0#V 0#V 0#V 0>Y#V 0>Y#V 0>Y#V 0#V 0#V 0#V 0%[#V 0%[#V 0%[#V 0%[#V0#V0#V0#V0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V/0#V0#V0#V0#V0#V 0QQ0c0c0c0c 0QQ0i 0QQ0l0l0l0l0l0l0l0l0l0l0l0l0l 0QQ0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u0u 0 00 0 0 0 0 0  00I0I0I0I@ 0I0I@ 0I0I0I 0 0ЋЋ0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00000000000( 00000( 00 0 0 0 0 0 0 0 0@ 0@ 0 @0@0@0(B 0mm@0@0@0(B 0mm@0X@0X@0X@0X@0X(B 0mm@0>@0>@0>@0>(B 0mm@0@0@0@0(B 0mm@0@0@0(B 0mm@0@0@0@0@0@0@0A 0YY@0B@0B@0B@ 0A 0@0;@0;@0;@0;@ 0;@ 0;@ 0;@0;@0;@0;@0;@0;A 0(B 0@0@0(B 0@0(B 0@0(B 0@0)(B 0@0(B 0@0:(B 0@0(B 0@0M@0M(B 0@0(B 0 @0G@0G@0G@0G(B 0 @0(B 0 @0(B 0 @0 @0 A 0@0[ @ 0@05A 055@0>@0>@0>@0>@:0>@90>@:0>@0>@:0>@0>@:0>@90>@:0>@90>@0>@:0>@0>@:0>@0>@:0>@0>@:0>@0>@:0>@0>@:0>@0>@0>@ 0>@ 0>@ 0>@ 0>@0>@0>@0>(B 0>>8@0m+@0+8@0m+@! 0-@! 0-@! 0-A! 0--A! 0--8@0m+@0B.@0B.8@0m+@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R1@0R18@0m+@0U8(B 0>>8@0:@0%:8@0:@ 0:@ 0:@ 0:@ 0:8@0:@0;@0;8@0:@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<@0g<8@0:@0>(B 0>>8@0e@@0y@8@0e@@ 0A@ 0A@ 0A@ 0A@ 0A8@0e@@0 B@0 B8@0e@@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B@0B8@0e@@0D(B 0>>8@0F@0F8@0F@ 0G@ 0G@ 0G@ 0G8@0F@0H@0H8@0F@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI@0NI8@0F@0dK(B 0>>8@0LM@0lM8@0LM@ 0N@ 0N@ 0N@ 0N8@0LM@0hO@0hO8@0LM@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP@0KP8@0LM@0)R(B 0>>8@0;S@0NS8@0;S@ 0|T@ 0|T@ 0|T@ 0|T@ 0|T@ 0|TA 0wU|TA 0wU|T8@0;S@0fV@0fV8@0;S@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y@0Y8@0;S@04[(B 0>>8@0]@0]8@0]@" 0s^@" 0s^@" 0s^8@0]@0L_@0L_8@0]@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`@0`8@0]@0b(B 0>>0@0e@05e0@0e@# 0Ch@# 0Ch@# 0Ch@# 0Ch@# 0Ch@# 0ChA# 0iChA# 0iCh0@0e@0&j@0&j0@0e@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l@0 l0@0e@0wm(B 0>>8@0n@0o8@0n@* 0ls@* 0ls@* 0ls@* 0ls@* 0ls8@0n@0;t@0;t8@0n@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx@0zx8@0n@0+z(B 0 >>8@0|@0$|8@0|@+ 00@+ 00@+ 00@+ 00@+ 00@+ 00@+ 00@+ 00@+ 00A+ 0h0A+ 0h0A+ 0h0A+ 0h0A+ 0h08@0|@0)@0)8@0|@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@08@0|@0`(B 0 >>8@02@0W8@02@, 0M@, 0M@, 0M@, 0M@, 0M@, 0MA, 0MA, 0M8@02@0=@0=8@02@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J@0J8@02@0+(B 0 >>8@0@08@0@- 0o@- 0o@0o8@0@0@08@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@08@0@0(B 0 >>8@0@08@0@. 0@. 0@. 0@. 0@. 0@. 0@. 0@. 0@. 0@. 0 @. 0 A. 0 A. 0 A. 0 A. 0 A. 0 A. 0 8@0@0 @0 8@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@08@0@0 (B 0 >>8@0¶@0ݶ8@0¶@/ 0@/ 0@/ 0@/ 0@/ 08@0¶@0:@0:8@0¶@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@08@0¶@0(B 0>>8@0y@08@0y@0 0@0 0@0 0@0 0@0 0@0 08@0y@0@08@0y@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@08@0y@0(B 0>>8@0'@0=@0=8@0'@8 0@8 0@8 0@8 0@8 0@8 0A8 0A8 08@0'@0@08@0'@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X@0X8@0'@0O@0O(B 0>>8@0@0@9 0@9 0@08@0@0@: 0@: 0@: 0@: 0@: 0@: 0@: 0@: 0@: 0@: 0 @: 0 @: 0 @: 0 8@0@08@0@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:@0:8@0@0\A 055(B 0ZZ@0j@0j@0j@0j@0j@0j(B 0ZZ@0@0@5 0@5 0@0@4 0@4 0@4 0@4 0@0@B 0@B 0@0@C 0@0@0@0@0@0@0@0@0@0@0@0(B 0ZZ@0$@0$@ 0@0@ 0@0 @0 @0 @0 @30 @30 @30 @30 @40 @ 0 @0?@0?@0?@0?@0?@ 0 @0@0@0@0@ 0 0/#;00%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0%0R@0@0@0@0@0@0@0@00Rc@&0P@0P@$0Q@$0P@0P@$0P@0P 00Yc 0_c:004:0 0:00:00؀:00H:00:0#0(:0+00!>0 """""""""""""""""""""""""""""""""%?   E `!Fn%~"H]EN"# !!`#r+>KZbfijpkl{mqx|Dc"Ӣ8oϯؼ=d#'u*d26Q>ByGL V_ejpqOtkT$fl`gP 5{_Vz"-1y47=DKPyTjj,/012345789:;<=>?@ABCDEFHIJKLMNOPQRSTUVWYZ[\]^`abdflq|   "$'.8CMWagmxG 0 ;Bqciij4jfjjjj k)kkkkkl0lJlelllllmm@mVm {+Ea{[aK k,B9999:F:s:::;);V;;;; <><s<<<<&=T===n@QCCCCDTDF7IlIIII4JL+PXPPPQT W6WkWr 7f&WBs7n 7fEr'XHu'TxG1^* Gra=M*X<zRjj-6GX_ceghijkmnoprstuvwxyz{}~   !#%&()*+,-/012345679:;<=>?@ABDEFGHI JKLNOPQRSTUVXYZ[\]^_`bcdefhijklnopqrstuvwyz{|}~j.=Mhjkm $&')Hp9;<>]t&ADEGf!$%'F]x{|~ !<?@Bar)DGHJi"5PSTVu ! $ % ' F n ' B E F H g     & = X [ \ ^ } % ( ) + J d   D _ b c e *-.0O#?Z]^` $'(*Ih9<=?^y=X[\^}!$%'Fm (C^abd*-.0Oe1LOPRq#8SVWYx$A\_`bVqtuw###&7&;&CXYXbXeX~XXXXX]]]g^^^_2_;_ccc1gIgRgpppdx|xx}}}x6NW~‚Ă5>Jac.7}Ζ Ә˙ԙ?VX1:09`xСҡ9Q[3NP^x~f~Ʋɲ.FPνҽ~2JT`wz%=GUmw_w8ORRkmo6@ ;RU )/GQ7OYg3/K/U/344{444888:::}@@@GGGMMMVVVM^d^g^```i%i(iiiiuuurxxx} @WZl"9<0:Ofis\svC^`'?I2JT!:<E\_ @WZ   hO X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%X%̕:                 :::                       :   %!!XpzV812@10(  B S  ?hO _Hlt11647789 _Toc7500854 _Toc7500855 _Toc11647534 _Toc7500856 _Toc11647535 _Ref531573279 _Toc11647536 _Ref530816124 _Toc11647537 _Ref530803878 _Ref8723579 _Ref530812131 _Ref530806740 _Toc11647538 _Ref10887901 _Toc11647539 _Ref8030558 _Toc11647540 _Ref531501615 _Toc11647541 _Ref530822123 _Toc7948960 _Ref530885112 _Toc11647542 _Toc11647543 _Ref531763254 _Ref531763258 _Toc11647544 _Ref531749592 _Toc11647545 _Toc11647546 _Ref11047897 _Ref10529456 _Ref531749173 _Ref530829785 _Toc11647547 _Ref530883055 _Toc11647548 _Ref530884156 _Toc11647549 _Ref530885345 _Ref530885349 _Toc11647550 _Ref530886540 _Toc11647551 _Ref10887249 _Ref531510835 _Toc11647552 _Ref530888667 _Toc11647553 _Ref530892005 _Ref530829811 _Toc11647554 _Ref530901010 _Ref530901197 _Toc9066576 _Toc11647555 _Ref530882302 _Toc11647556 _Toc11647557 _Ref531761766 _Ref531762011 _Toc11647558 _Toc11647559 _Toc11647560 _Toc11647561 _Toc11647562 _Toc11647563 _Toc11647564 _Toc11647565 _Toc11647566 _Ref11560948 _Toc11647567 _Toc11647568 _Toc11647569 _Toc11647570 _Toc11647571 _Toc11647572 _Ref530882316 _Toc11647573 _Toc11647574 _Toc11647575 _Ref531402157 _Toc11647576 _Ref531403222 _Toc11647577 _Toc11647578 _Toc11647579 _Toc11647580 _Toc11647581 _Ref531421544 _Toc11647582 _Ref531488842 _Toc11647583 _Toc11647584 _Toc11647585 _Toc11647586 _Ref7509864 _Toc11647587 _Ref531510849 _Toc11647588 _Ref531504567 _Toc11647589 _Toc11647590 aggregation _Ref7509809 _Toc11647591 _Ref531665805 _Ref531665809 _Ref530886833 OrderModel _Toc11647592 _Toc11647593 _Ref536603665 _Ref536603733 _Toc11647594 _Ref536603861 _Ref536603888 _Toc11647595 _Toc11647596 _Toc11647597 _Ref8111122 _Ref8029913 _Toc11647598 _Toc11647599 _Toc11647600))QQ#V#VdXXccciilluu}   HCϋvޔ8ջY1%%4}gR     {'36:f<BMI9XP iO@  !"#$%&'(*)+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmonpqrstuvwxyz{|}~&))QQEVEVXXcc diimmvv} Jłqqd ߔYӡʲ {S%Q( V # 5 5  '4%6:y<BlINOYh^5a)i kxÇ>>G~~w:`[mZ'' !iO# $ D% |IJMKTL M|N?OPwQwD?П<RdSTUdV$WXYdZ$[\ܟݟdޟ$]~^~d~_$~`}a}bd}$}L c̐deLf ̏ghL i,jkl,mnolp (q,rst}uT}v}w|x|yT|z|{{|{}+~|+<+++|+<+++~~T~<DbD, $x-# <Ϳ    4$,LqL4dl lO#|LڿD,mT !e !"4"#L$<%&"'(4̼)!*L+D,-ҽ./D 0L1 23D454"6D7<}#8N9c:!;}#<= >( ?@dA,BCDEF4 Gd !H4 <  $!.=O556::AA AMMNNNNeӃBBU   c::##   Z$Z$d$$$$$% % %%%%%%&&&`'`'n'S)7*7*A* . .2.2."/"/*///1111(141`122v3v3~3324444466 667777777788 888+8+838m9m9x99:: ; ;;/</<8<D<<<<=====>>>[?[?j?@@@AAABBBBBEE,G,G6GQG%H%H.HHHHiO      !"#$%&'()*+,-./1023465789;:<=>?@ACBDEGFHIJKLNMOPQRSVTUWXYZ\[]^_a`bcedfgihjklmnoqprstvuwzxy{|}~;Ic556::AAANNNNNNqǃ߃GGX   c??((   b$j$j$$$$%%%%%%%%%&&&l't't'\)?*D*D*,.,.5.5.(/0/0///11&1217171m122|3333444444666677777777 888#8#8185858v9999::;;;6<B<G<G<<<<=====>>>h?o?o?@@@AAABBBBBEE4G?@ACBDEGFHIJKLNMOPQRSUVTWXYZ\[]^_a`bcedfgihjklmnoqprstvuwyzx{|}~ >H*urn:schemas-microsoft-com:office:smarttags PostalCode9*urn:schemas-microsoft-com:office:smarttagsState8*urn:schemas-microsoft-com:office:smarttagsCity:L*urn:schemas-microsoft-com:office:smarttagsStreet=A*urn:schemas-microsoft-com:office:smarttags PlaceType=@*urn:schemas-microsoft-com:office:smarttags PlaceNameB*urn:schemas-microsoft-com:office:smarttagscountry-region9*urn:schemas-microsoft-com:office:smarttagsplaceV*urn:schemas-microsoft-com:office:smarttags PersonNamehttp://www.microsoft.com;M*urn:schemas-microsoft-com:office:smarttagsaddress8N*urn:schemas-microsoft-com:office:smarttagsdate 1120026DayMonthYearNMLHj+q+GGJJJJB`G`M`U```````````aa!a)a2a?a~aaaaaabbIbNbYb\bbbbbbbbbccWcZckcvcee,f.f?fDfffffggggggggOqqqrrArCr_r`rrs*s,s;s>sXsYsrsssssss wwwUwwwww4x;xADAAAAALBQBCCFFGGHIKKMMOMNNNNO#O7P:P:Q>QQQQQRRYRbRRRRRfVpVX XXXYZ#[*[]]__ ````bbbbbbbb1c3ccc e e>eEeeeeeeeeepfufiijj'{+{y||||!}&}(}2}[}i}}}}} ~~-~;~w~~~~~'35EWgg{'2N\ $W\~Y\+PZ¦Φ$ר~(2ǫ:Aϳ۳4EFHISTVWbdrsuvضٶ۶ܶ "3467ABDEPR^_bcnors~ͷηӷԷ  '(-.;<ABOQ\]bcpqvwʸ˸иѸ޸߸+-89>?LMRS`brsuvƹȹչֹ۹ܹ  ')78<=ABIJNO[]klpq}~ӺԺֺ׺!./12<=?@ACJLZ[_`mnrs»ûȻɻֻ׻ܻݻ    '(,-9:>?KMSzڼ6;ivwz{ɽ˽ڽ۽ݽ޽  *,89;<FGIJTVcdghstwx¾þǾȾԾվپھ &')*4578CEPQVWdejkxzÿĿƿǿѿҿԿտ߿0134>?ABLN\]_`jkmnxKR^e2>.=>@AKLNOY[fgjkvwz{ )*,-79DEIJVW[\hjuvz{  *,78<=ACJKOP]_mnpq{|~ !+,./9;HILMXY\]hjyz|}UZbj`~  !!$$`'f'Q(S(******+++C,,--!-2233^7f77778n8R<W<>>/B6BBBBBDDDDDEEEGGiGmGHKKKK+ONOOOiO,'1'1 1CCEEB`G`M`U```````````aa!a)a2a?a~aaaaaaaabb7b@bIbNbsb|bbbbbbbccWcZckcvcDeIef"fgg:mDmNqqqrr^r_rrsKLZ[adjkxyжѶ޶߶  "+,9:HIORVWfgvw}ŷƷٷڷ 34GHNQTUhi|}¸øָ׸ #$*-01DEXY_bjkxyŹȹ͹ι &)/0ABSTZ]cduv˺̺ٺں!&'45BCILRSefwx~λϻ 12CDJino½Ƚ˽ҽӽ"#),01>?LMSV[\kl{|̾;޾߾  ,-;<BEHI\]pqwzɿʿ׿ؿ޿ ()67DEKNTUbcpqw.56CDQRX[^_no~ !"/069<=NO`agjmn"#),/0BCUV\_efst #$128;@APQ`agjqr/3oEM # #####$Y$l$r$s$$$%% % %)%*%n%o%%%%%&&H&I&&&&&&'_'v'{'|''''(*(+(5(Q(S(U(o()R)c)i)j)))6*F*_***+++*+~+++++E,_,r,,,,,!-"-U---7.t..!/0/2/G////0,00008191A1B1n1111112B3333#4&4'4C4K4N4y44444 5 5M5N555666,6n6o666666I7S7T7\7t77777786878n888889A9B9999::,:3:T:e:m:o:: ; ;5;6;p;I<Q<<<<=5========4>>>>>>>> ?Z?q?y??@@@@@AB8BBBBBBCCC C`CCCCCDDHDIDcDdDnDoDDDDDDDEEEFxFFFF+G>GFGGGnGGHH8HHKK+ONOOOiO:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::'J2UE`@[  MHHTIIIJJJJJJKKKKdMeMjMlMmMN*O*O+O3OMOOOiOH+ONOOOiOSimon Corston-Oliver Michael Gamon Eric RinggerC 3KH j=hGA=h #_.v_ZCe@8 tfW^ bG_3 Йr%3 p~9> py4l3w(l@O]arЋ:3h.8%NJ xLЙr"'[t^u? 7!Du+H0"|F3|#uY$džQ<]$R9z'.d(R#q!E,Z$AZ,~" ]-V y~-0SC .0D/Jb"i52^F2R3h.83dv7"*:zXD2>(/uNEG uM+~K<=N5LPDPmZ ) \M\džv:cbj=eR;#f:njyQ^k̚Ikې1|l(3m־`Lo5K:p;> tl@t,@eu?u\| }NJ8Tu}K 4~bh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hH P^`PhH @@^@`hH. 0^`0hH.. ``^``hH... ^`hH .... ^`hH ..... ^`hH ......  `^``hH.......  00^0`hH........ P^`PhH @@^@`hH. 0^`0hH.. ``^``hH... ^`hH .... ^`hH ..... ^`hH ......  `^``hH.......  00^0`hH........h88^8`OJQJo(hHoh^`OJQJ^Jo(hHoh  ^ `OJ QJ o(hHh  ^ `OJQJo(hHhxx^x`OJQJ^Jo(hHohHH^H`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh88^8`OJ QJ o(hHh ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh88^8`OJ QJ o(hHh ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJQJo(hHh^`OJQJo(hHhpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHh^`OJ QJ o(hHhpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHhhh^h`OJ QJ o(hHh88^8`OJQJo(hHoh^`OJ QJ o(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHohxx^x`OJ QJ o(hHhHH^H`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh88^8`OJ QJ o(hHh ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh88^8`OJ QJ o(hHh ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh88^8`OJ QJ o(hHh ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`duled 1mal2Michael Gamon, Eric Ringger, Simon Corston-Oliver0OGerman; sentence realization; natural language generation; machine translationerm normal.dottSimon Corston-Oliveriza106Microsoft Word 10.0@@0@F@!ޓ@ŊGIJW  FMicrosoft Word Document MSWordDocWord.Document.89q OJQJ^Jo(hHohPP^P`OJ QJ o(hHhhh^h`OJ QJ o(hHh88^8`OJQJo(hHoh^`OJ QJ o(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHohxx^x`OJ QJ o(hHhHH^H`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHhhh^h`OJ QJ o(hHh^`OJQJo(hHoh^`OJ QJ o(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHohxx^x`OJ QJ o(hHhHH^H`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJ^Jo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hH՜.+,D՜.+,h$ hp  Microsoft ResearchT%*G -Amalgam: A machine-learned generation module Title< 8@ _PID_HLINKSA0 _Toc116476009 _Toc116475999 _Toc116475989| _Toc116475979v _Toc116475969p _Toc116475959j _Toc116475949d _Toc116475939^ _Toc116475929X _Toc116475919R _Toc116475908L _Toc116475898F _Toc116475888@ _Toc116475878: _Toc1164758684 _Toc116475858. _Toc116475848( _Toc116475838" _Toc116475828 _Toc116475818 _Toc116475807 _Toc116475797  _Toc116475787 _Toc116475777 _Toc116475767 _Toc116475757 _Toc116475747 _Toc116475737 _Toc116475727 _Toc116475717 _Toc116475706 _Toc116475696 _Toc116475686 _Toc116475676 _Toc116475666 _Toc116475656 _Toc116475646 _Toc116475636 _Toc116475626 _Toc116475616 _Toc116475605 _Toc116475595 _Toc116475585 _Toc116475575 _Toc116475565 _Toc116475555z _Toc116475545t _Toc116475535n _Toc116475525h _Toc116475515b _Toc116475504\ _Toc116475494V _Toc116475484P _Toc116475474J _Toc116475464D _Toc116475454> _Toc1164754448 _Toc1164754342 _Toc116475424, _Toc116475414& _Toc116475403  _Toc116475393 _Toc116475383 _Toc116475373 _Toc116475363 _Toc116475353 _Toc11647534h^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hH P^`PhH @@^@`hH. 0^`0hH.. ``^``hH... ^`hH .... ^`hH ..... ^`hH ......  `^``hH.......  00^0`hH........h^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hH hh^h`hH) ^`hH) 88^8`hH) ^`hH() ^`hH() pp^p`hH()   ^ `hH. @ @ ^@ `hH.   ^ `hH.h^`OJQJo(hHh^`OJ QJ o(hHhpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHhhh^h`OJ QJ o(hHh88^8`OJQJo(hHoh^`OJ QJ o(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHohxx^x`OJ QJ o(hHhHH^H`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh                  ^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hH^`OJPJQJ^Jo(-^`OJQJ^Jo(hHopp^p`OJ QJ o(hH@ @ ^@ `OJQJo(hH^`OJQJ^Jo(hHo^`OJ QJ o(hH^`OJQJo(hH^`OJQJ^Jo(hHoPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHhhh^h`OJ QJ o(hHh88^8`OJQJo(hHoh^`OJ QJ o(hHh  ^ `OJQJo(hHh  ^ `OJQJ^Jo(hHohxx^x`OJ QJ o(hHhHH^H`OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh 88^8`hH.h ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh88^8`OJ QJ o(hHh ^`hH.h  L ^ `LhH.h   ^ `hH.h xx^x`hH.h HLH^H`LhH.h ^`hH.h ^`hH.h L^`LhH.h^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh ^`o(hH.h^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh ^`hH.h ^`hH.h pLp^p`LhH.h @ @ ^@ `hH.h ^`hH.h L^`LhH.h ^`hH.h ^`hH.h PLP^P`LhH.h^`OJQJo(hHoh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJQJo(hHoh^`OJQJo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHh^`OJ QJ o(hHh^`OJQJ^Jo(hHohpp^p`OJ QJ o(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJ QJ o(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJ QJ o(hHEGAF2YB F2ġB F2B 5LPpyt[0Ik07!0.v0_3 0 3Lo0D2>|lO]C .y~- xL=eD/8%mv:cK:p"i5283P }uNE08Tu}uM^ ( tz'0u\|G09>#q!E,eunjH0";#f ]-:3mZR3%3 0r|#3wAZ,0K<=Ndv7) \u? uY$0M\Q<]$0*:.d(Q^ke@8 4~ jZB )B) 3r[{CdK,\B x[!W|Wx[!:.*x[!yB-K,\LG-x[!xh/3r[kg5!(5K,\/P;x[!FD3r[ I3r[P'X3r[6`XK,\IOZ3r[K,\;&eK,\9ag3r[ceJjx[!V|sjx[!jK,\U$p (~pdE[]rIM T: o  7 i; UG B9(+dVKxAf*d:pgg}bS NTnbK[Yvufs q"X2+ "y56P  .I R ! !B!F!Z!d! ";h#lJ$k$&%(&y:&'B'uL'A(+U+k+O ,D,jR,0U,6e,@m,X*-5-P-[v-3.gz.{.$/'/71D1Rp1 23t36+5"6e%6R7Yh7nk7y7x87k9v9,;3;#D;P<V<c(=@ >!> T>T>Y?. A3B&C?ECkiDEE-ECEkEkEoE;FJF,/FGHWaHcI()I,K/GKxK{KL{L"LeWL]:MSMC\MjMNFNVOc6OgOiO P!P.PlQyQ RReR STT%TdTm#UX9UG:ULXUVSVyoWa5XY.;YbjY~Y ZZ#6ZS^Z[9[0[&\G-\3E\R\N4]J]Z]i]j]J^_^_2_3_N<_ `'`H`q`aa$ama !bnXdwde.ep6eDkfgh/gDKgTghh%hyheiI j=jkijk l.l:nl|l m&m(mVEmo8oMoo{pXp3p[Npdp$ri>sQsbsdgtptv3vUv'wQw LxrsxvxeyPy,ym2yqy+z}z{ {8,{:5{VU|~|k}~.~+\veN8R shcXU6U6/kHMEfg4NwAz- %'|?sh7c@u -#w~DR\2gb+gs_f' a=#OO[E*'H%T:|"vF\7}21>RT1R$.? U{VKY800 m:?[f,lc5s\'A["X *CT5192EZ$1n9Tz3n3F*|bwkF{Rt} :O( A!iM&:0+4wW/\'e-2WD6[^Dqy U(&FN/]~8Krl_$6E)0ciq&:iEv1I&V_I4w;Oh)!X3v]5wuWY'4P!|%)Z8NpTcgj& ]gOX B>xMxz37bzAo"j4]8\7y'WklDgiP- (!.b&f/^QJJJJpMesyԵյBFKLim#8S@IL*1?"(,5\`r˿Ͽؿ _f{!V)Q=LYgFzQ@bk+8 9 B    M     )7< $'#/ R]p{#P'W'l'y'-U/////////////////00*0102080H0X0_0`0f0r00000000000000001111.1>1E1F1K1W1c1j1k1p1111111111111111 2222%252E2L2M2P2d2x222222222222222233 33-3435383L3`3g3h3n3~333333333333334V448s9w999999999999999::":):*:1:D:W:^:_:h:{::::: ><>>>Y?]?g?n?x?y?????????????@ @ @@%@9@@@A@F@Z@n@u@v@DJDOErEMFQF[FbFlFmFrFFFFFFFFFFFFFFFvGLLoL-M1M;MBMLMMMPM`MpMwMxM}MMMMMNgRUUFVJVTV[VeVfVqVVVVVVVVVVM[[\]]]]]]]]]]]]]^^^^!^0^?^E^F^h^^`hhhhhhhhhhhhhhii)iuiixput.u2uL[bckyѶ߶  !,:IPQWgw~Ʒڷ 4HOPUi}ø׸$+,1EY`akyƹǹι '(0BT[\dv̺ں  '5CJKSfxϻ 2DKLTbqxyѼؼټ -45<N`gho½ɽʽӽ#*+1?MTU\l|;߾ -<CDI]qxyʿؿ߿)7ELMUcqxy2"E%,-6DRYZ_o  "078=Oahin#*+0CV]^ft$29:AQahirdS&3@GH&-78?Ukrs{189@Rdkl.2<CMNUk!");MTUwHRIiO-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0KKK-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-0-05a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a5a-0-0-0-0-0-0-0-0-0-0555I0a0a0a0-0-0-0uv0a0a0C9-0@77J;~~77LA+A,78LhO`@`4`l@`B`@` @UnknownSimon Corston-Oliver&Simon Corston-Oliver20020606T1624233774f EmailStyle17CNSimon Corston-Oliver Sz Times New RomanTimes[SymbolBookshelf Symbol 3G& z ArialHelvetica3: TimesY5  z Courier NewArial NarrowASimSun?????M& z!TahomaArial BlackQ5  hMS Mincho?l?r ??fc71 CourierI WingdingsSymbol"qhafcffcffjJWI*%JWI*%1>4dGG 3q)?8,{,Amalgam: A machine-learned generation moduleNGerman; sentence realization; natural language generation; machine translation1Michael Gamon, Eric Ringger, Simon Corston-OliverSimon Corston-OliverC                           ! " # $ % & ' ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < = > ? @ A B