ࡱ> {` _bjbjFF ,,%>j j j j nnnD####D$DW*%.X("z(z(z(,I/%0pjKlKlKlK%KyQaW$Yh\Wn_8+,_8_8Wj j z(z(qWAAA_8Xj z(nz(jKA_8jKAAV4D@$JnDz(% #;TtD .K<W0WD \ =<\DDJ\nDX0#2A?3#4<000WWGAd000W_8_8_8_8d##j j j j j j  Investigations of Topic Dynamics in Web Search Xuehua Shen Department of Computer Science University of Illinois Urbana, IL 61801 +1 217-244-1036 xshen@cs.uiuc.eduSusan Dumais Microsoft Research One Microsoft Way Redmond, WA 98052 +1 425-706-8049 sdumais@microsoft.comEric Horvitz Microsoft Research One Microsoft Way Redmond, WA 98052 +1 425-706-2127 horvitz@microsoft.com ABSTRACT We report on an investigation of transitions among the topics of pages visited by a sample population of users of MSN Search. We learn probabilistic models of topic transitions for individual users and groups of users. In an effort to compare topic transitions for individuals versus larger groups, we consider the relative accuracies of personal models of topic dynamics with models constructed from sets of pages drawn from similar groups and from a larger population of users. To explore temporal dynamics, we compare the accuracy of these models for predicting transitions in the topics of visits at increasingly more distant times in the future. Finally, we touch on promising directions for applying models of search topic dynamics. Categories and Subject Descriptors H.3 Information Storage and Retrieval General Terms Algorithms, Measurement, Experimentation. Keywords Web search, log analysis, topic analysis, open directory project, web mining, predicting user behavior INTRODUCTION The Web provides opportunities for gathering and analyzing large data sets that reflect users interactions with web-based services. Analysis and synthesis of the rich data provided by these logs promises to lead to insights about user goals, the development of techniques that provide higher-quality search results based on enhanced content selection and ranking algorithms, and new forms of search personalization. We describe research that examines characteristics of the topics and transitions among topics associated with page visits by users engaged in searching on the Web. We construct probabilistic models to characterize the distribution of topics for individuals and groups of users. We construct the predictive models using a training corpus of tagged pages, and then use these models to predict the topics of subsequent pages by users. Copyright is held by the author/owner(s). WWW 2005, May 10--14, 2005, Chiba, Japan. To probe the differences between the predictive power of personalized models and the models built by analyzing groups of users, we perform several comparative studies. We construct Markov and marginal models with data drawn from (1) single individuals, (2) composite data from people who have the same topic dominance in the pages they visit during their search sessions, and (3) data from the entire population of users. For these different classes of models, we perform temporal studies that consider the predictive accuracy with increasing periods of time between page visits used in training the models and evaluating the accuracy of the learned models. Finally, we discuss several applications of models of topic dynamics. RELATED WORK The ability to model and predict users search and browsing behaviors has been explored by researchers in several areas. The analysis of URL access patterns has been used to improve Web cache performance (Davison  REF _Ref87664034 \r \h  \* MERGEFORMAT [6], Deshpande and Karypis  REF _Ref87664304 \r \h  \* MERGEFORMAT [7], Lempel and Moran  REF _Ref87664054 \r \h  \* MERGEFORMAT [14], Schechter et al.  REF _Ref87664084 \r \h  \* MERGEFORMAT [22]) and to guide prefetching (Horvitz  REF _Ref87664252 \r \h  \* MERGEFORMAT [10]). In general, models developed for caching and prefetching average over large numbers of users, and exploit the consistency in access patterns for individual URLs or sites, but do not consider topical consistency. Another line of research has explored the paths that users take in browsing and searching web sites. Pitkow and Pirolli  REF _Ref87664886 \r \h  \* MERGEFORMAT [17] and Chi et al.  REF _Ref87664900 \r \h  \* MERGEFORMAT [4] used clustering techniques to group users with similar access patterns, with the goal of identifying common user needs. This work involves detailed analysis of individual web sites. There has been some recent work exploring how page importance computations like PageRank can be specialized to different topics (Haveliwala  REF _Ref87665740 \r \h  \* MERGEFORMAT [9]) or to query-specific results (Richardson and Domingos  REF _Ref87694830 \r \h  \* MERGEFORMAT [19]). In Haveliwalas work, pre-computed hub vectors that correspond to broad topical categories were used to generate different page importance scores for different topics. The focus of this work is on algorithmic techniques rather than the evaluation of predictive accuracy or personalized search applications. There is a large body of work on constructing user profiles based on explicit profile specification or on the automatic analysis of the content and link structure of Web pages visited (Ravindran and Gauch  REF _Ref87665524 \r \h  \* MERGEFORMAT [18], Sarukkai  REF _Ref87665539 \r \h  \* MERGEFORMAT [21], Sugiyama et al.  REF _Ref87665552 \r \h  \* MERGEFORMAT [25]). In general, this work develops models for individual searchers and does not explore group models or the evolution of interests over time. Several groups have examined Web search behaviors by analyzing Web query logs  REF _Ref87424026 \r \h  \* MERGEFORMAT [1] REF _Ref87515274 \r \h  \* MERGEFORMAT [2] REF _Ref87515276 \r \h  \* MERGEFORMAT [20] REF _Ref20533750 \r \h  \* MERGEFORMAT [23] REF _Ref20533752 \r \h  \* MERGEFORMAT [24]. Broder  REF _Ref87515274 \r \h  \* MERGEFORMAT [2] and Rose and Levinson  REF _Ref87515276 \r \h  \* MERGEFORMAT [20] characterized different information needs that users have in searching. They describe searchers as motivated by navigational (getting to a web page), informational (learn sometime about a topic), transactional (buy something) or resource (obtain something or interact with someone) goals. Topic or content is largely orthogonal to information needs. For example, searchers want to buy things or find out information about a variety of different topics (arts, computers, health, sports). Silverstein et al.  REF _Ref20533750 \r \h  \* MERGEFORMAT [23] and Spink et al.  REF _Ref20533752 \r \h  \* MERGEFORMAT [24] have analyzed large query logs and summarized general characteristics of Web searches, including the length, syntactic characteristics and frequencies of queries, the number or results pages viewed, and the nature of search sessions. A few research groups have explored the diversity and dynamics of topics that people search for. Spink et al.  REF _Ref20533752 \r \h  \* MERGEFORMAT [24] asked human judges to label a small sample of 2,414 queries. Judges assigned each query to one of 11 topical categories. The most common categories observed in their sample were Entertainment, Adult and Commerce. Lau and Horvitz  REF _Ref87516196 \r \h  \* MERGEFORMAT [13] examined topical distinctions using a sample of 4,690 hand-labeled queries that were assigned to a broad ontology of 15 informational goals. The most common topics in their study were Products and Services, Adult and Entertainment. In addition to summarizing general distributional characteristics of queries, Lau and Horvitz also analyzed the temporal dynamics of query reformulation strategies. A Bayesian model was constructed which related variables such as the topic of the current query, the refinement relationship to previous queries, and the time between adjacent queries. The model could be used to predict topics, dwell times, and query refinements. Beitzel et al.  REF _Ref87424026 \r \h  \* MERGEFORMAT [1] recently reported on a much larger scale analysis of query topics and dynamics. They analyzed all the queries submitted to a commercial search service over a one week period of time. Queries were automatically assigned to one of 14 general topical categories by matching the queries to lists of terms corresponding to each category. The lists were manually constructed by human editors. About 13% of the queries could be assigned by this technique, and this represents millions of user queries. The most common categories were Adult, Entertainment and Music. They also explored difference in the distribution of topics over the course of the day, finding that Adult queries were most common in the early morning hours and Personal Finance queries peaked just before noon. All of their data summaries were aggregated over all users in their sample. Our contributions include examining topic dynamics over a long period of time (5 weeks) with a large number of users. Instead of inferring the topic of interest using the query which is often very short and ambiguous, we identified the topics associated with URLs that individuals visited. In addition, we characterize the predictive power of individual models versus models built for large groups of users. And, we also considered the influence of differences in time between when the topic models are constructed and when they are evaluated on new data. MODELING TOPIC DYNAMICS Our goal is to understand users search behaviors by analyzing log data from a large number of users over an extended period of time. As described in more detail below, we start with a large log of queries and URLs visited over a period of five weeks. Each URL has a topical category (e.g., Arts, Business, Computers, etc.) associated with it. We wish to understand the nature of topics that users explore, the consistency of the topics a user visits over time, and the similarity of users to each other, to groups of users, and to the population as a whole. Beyond elucidation of topic dynamics from large-scale log analysis, we believe that better understanding of the dynamics of topic viewing over time will allow us to better interpret queries and identify informational goals, and, ultimately, to better personalize search and information access. In the rest of the paper we construct probabilistic models of the pages visited by individuals, groups of individual and the population of users as a whole. We report basic statistics about the number of topics that individuals explore, and topic dynamics as a function of time. The main focus of our experiments is to predict the topic of each URL that an individual visits over time. We use different techniques to predict the topics of URLs based on marginal topic distributions and Markov transition probabilities. We use models derived from analyzing the patterns observed in individuals, groups of similar individuals, and the population as a whole. Models Marginal Models The marginal models simply use the overall probability distribution for each of the 15 topics. The marginal models serve as a baseline for richer Markov models. Markov Models The Markov models explicitly represent the probabilities of transitioning among topics. That is, we consider the probability of moving from one topic to another on successive URL visits. The model has 225 states, each representing transitions from topic to topic (including transitions to the same topic). Time-specific Markov Models The time-specific Markov models are a refinement of the general Markov model. Again, we estimate the probability of moving from one topic to another, but use different models depending on temporal parameters. In one case, we simply vary the time gap between when the model is built and when it is evaluated. In another case, we build separate transition matrices for small time intervals (less than 5 minutes) and long time intervals (5 or more minutes) between successive actions to differentiate different topic patterns based on time interval. We use maximum likelihood techniques to estimate all model parameters, and Jelinek-Mercer smoothing to estimate the probability distributions  REF _Ref87698646 \r \h [11]. User Groups We construct models for individuals and for groups, developing marginal and Markov models for individuals, similar groups, and the population as a whole. We use these models to predict the behavior of individual users. Individual This technique uses the previous behavior of each individual to predict their current behavior. We suspected a priori that this would be the most accurate method, but it requires a large amount of storage and, as we discovered, appears to have data scarcity problems for the more complex models. Groups This technique uses data from groups of similar individuals to predict the current behavior of an individual. There are many techniques for defining groups of similar individuals. For the experiments reported here, we grouped together all individuals who had the same maximally visited topic based on their marginal model. Population This technique uses data from the entire population to predict the current behavior of an individual. EVALUATION Data Set Basic Characteristics The basic data consists of a sample of instrumented traffic collected from MSN Search over a five week period from May 22 to June 29, 2004. The instrumentation captured user queries, the list of search results that were returned, and the URLs visited from the search results page (but not pages that are viewed after this, since that would have required client-side instrumentation). The basic user actions we worked with were: Client ID, TimeStamp, Action (Query, Clicked), and Value (a string for Query, a URL for Clicked). The data in our sample includes more than 87 million actions from 2.7 million unique users. Queries accounted for 58% of the actions and URL visits for 42% of the actions. Client ID is identified using cookies, and no personally identifiable information was collected. There is certainly some noise inherent in identifying individuals using cookies (as opposed to requiring a login). However, this represents an important analysis scenario for search engine providers, and is the one we model in our work. Because we have been interested in exploring query and topic dynamics over time, we selected a sample of 6,153 users who had more than 100 actions (either queries or URL visits) over the first two weeks. This data set contains more than 660,000 URL visits for which we could assign topics (as described in detail in the next section) over the five week period. Topic Categories There are a number of ways to tag the content of URLs. We used topics from the Open Directory Project (ODP)  REF _Ref87261790 \r \h  \* MERGEFORMAT [16]. The ODP is human-edited directory of the Web, which is constructed and maintained by a large group of volunteer editors. The directory contains more than 4 million Web pages which are organized into more than 500,000 categories. For this experiment we used only the first-level categories from ODP. Our method works at any level of analysis, but we focused on top-level categories to make comparisons with earlier work easier. We omitted the Regional and World categories since we were interested in topical distinctions, and added an Adult category. The categories we used are: Adult, Arts, Business, Computers, Games, Health, Home, Kids and Teens, News, Recreation, Reference, Science, Shopping, Society and Sports. Category tags were automatically assigned to each URL using a combination of direct lookup in the ODP (for URLs that were in the directory) and heuristics about the distribution of categories for the site and sub-site of a URL (for URLs that were not in the directory). This technique is very quick to apply and gave about 50% coverage for the URLs clicked on. In our analysis of topic dynamics, we ignored URLs that could not be assigned a category tag in this manner described above. We studied the labels assigned to more than one hundred individuals in detail and did not detect any systematic bias in the URLs that were automatically assigned labels and those that were not. As described in more detail below, we are currently exploring techniques for improving the coverage of automatic topic assignment for URLs and for incorporating the query into topic assignment. One or more topics could be assigned to each URL. On average, we found that there were 1.30 second-level and 1.11 first-level topics assigned to each URL. Sample Logs Tables 1a and 1b show samples from the logs of two individuals. For each action, we show the Elapsed Time (in seconds since May 22, 2004 when our data collection started), the Action (query (Q) or click through on a URL (C)), the Value of the action (the query string or the clicked URL), and the automatically assigned First-level Categories (labeled TopCat1 and TopCat2). We include both queries and URLs in these samples to provide context, but only URLs were analyzed in developing our topic models. The individual in Table 1a asks a number of different questions over a five week period, but most are in the general area of computers and computer games. The individual in Table 1b shows much more variability in topics, including queries about arts, business, reference and health. Topic Prediction The main focus of our experiments was to predict the topic of the next URL that an individual will visit over time. Models were built using a subset of the data for training (e.g., data from week 1) and used to predict the remaining data (e.g., data from weeks 2-5). As outlined above, the main variables we explored were the type of model (Marginal, Markov, or Time-Specific Markov), and the cohort group used to estimate the topic probabilities (an Individual, a Group of similar individuals, or the entire Population). We also varied the amount of training data used to build models and temporal characteristics of the training set. We computed several measures for comparing the differences between two topic distributions. We measured the Kullback-Leibler (KL) divergence between the two distributions  REF _Ref87423664 \r \h  \* MERGEFORMAT [5]. The KL divergence is a classic information-theoretic measure of the asymmetric difference between two distributions. We also computed the Jensen-Shannon (JS) divergence which is a symmetric variant of the KL divergence  REF _Ref87423664 \r \h  \* MERGEFORMAT [5]. We also measured the predictive accuracy of the models in two different ways. The first approach computes a single score for each URL based on the overlap between the actual topic categories and the predicted topic categories. The second approach measures the accuracy of predicting each category, as is done in text classification experiments. We used the F1 measure, which is the harmonic mean of precision and recall, where precision is the ratio of correct positives to predicted positives and recall is the ratio of correct positives to true positives. Results from all the measures are in general agreement. Below we present results of the F1 accuracy measure, because it is widely used in the text classification literature  REF _Ref87498122 \r \h  \* MERGEFORMAT [15] REF _Ref87498280 \r \h  \* MERGEFORMAT [26]. In all cases, we built a model based on some training data and evaluate the model on a holdout set of testing data. For each test URL, we predicted which of the topics it belongs to. Each URL can be associated with zero, one or more than one topics. We compared these model predictions with the true category assignments generated by the automatic procedure described in section  REF _Ref87501283 \r \h  \* MERGEFORMAT 4.1.2. We report the micro-averaged F1 measure, which give equal weight to the accuracy for each URL. RESULTS General Data Characteristics We explored several characteristics of the dataset before developing the topic prediction models. For each individual we computed the total number of actions, queries, clicks, different topics and topic shifts. The queries and session behaviors of the users in this sample are similar to users in studies of query logs  REF _Ref20533750 \r \h  \* MERGEFORMAT [23] REF _Ref20533752 \r \h  \* MERGEFORMAT [24] in many respects. We found that queries were 2.43 words long on average. There were 2.49 actions per session (including both queries and URLs visited), where a session boundary is defined as a gap of more than 15 minutes between successive actions. For the 6153 users that we studied in detail, we found that the average number of different topics represented in the URLs selected by an individual was 7.2 with a standard deviation of 2.1. Very few individuals focused exclusively on one topic, and very few covered the full range of 15 topics over the five week period. Not surprisingly, the distribution of different topics is non-uniform. The three most frequent topics are: Arts (16%), Shopping (15%) and Society (12%). Similar topics have been reported in previous query log analysis  REF _Ref87424026 \r \h  \* MERGEFORMAT [1] REF _Ref87516196 \r \h  \* MERGEFORMAT [13] REF _Ref20533752 \r \h  \* MERGEFORMAT [24], although it is difficult to compare precisely because the categories used are not the same. Recall that the three most common topics in previous studies were: Spink et al.  REF _Ref20533752 \r \h  \* MERGEFORMAT [24] Entertainment, Adult Content and Commerce; Beitzel et al.  REF _Ref87424026 \r \h  \* MERGEFORMAT [1] Adult, Entertainment and Music; and Lau and Horvitz  REF _Ref87516196 \r \h  \* MERGEFORMAT [13] Products and Services, Adult and Entertainment. Our results on topic distributions were relatively consistent with the prior work, except that we see significantly less adult content than reported in these studies. The lack of adult content in our URL analysis is because the search engine runs with an adult filter on by default, so not much adult content is returned in response to searches. Table 2 summarizes the transitions from one topic to another. The rows represent the starting state and the columns the destination topic. The values are normalized by row, so that sum of transitions from one state to all other states is equal to 1. The bold numbers represent the most common transition in each row. In general, transitions from a state to itself are the most common. There are some cases where transitions to the most common state (Arts) are higher than self transitions. (This is the w1 Population Markov model, described in more detail in the next section.) We now turn to our analyses of the accuracy of a variety of different models in predicting the topics of URLs visited by individuals. Marginal and Markov Models Figure 1 shows the accuracy for topic predictions for the Marginal and Markov models, and for each group of users (Individual, Group and Population). For the data reported here we used week 1 (w1) data to train the models and evaluated the models on week 2 data (w2).  For the Marginal model, topic predictions are most accurate when using the Individual and Group models. The similar performance of the Individual and Group models reflects the fact that we grouped users based on the maximum topic in week 1. The advantage of the Individual and Group models over the population models shows that users are consistent in the distribution of topics they visit from week 1 to week 2. Prediction accuracy is consistently higher with the Markov model than with the Marginal model for all groups. This shows that knowing the context of the previous topic helps predict the next topic. For the Markov model, topic predictions are most accurate with the Group and Population models. We believe that the relatively poor performance of the Individual Markov model is a result of data sparcity, because many of the topic-topic transitions are not observed in the training period. If we look at the self-prediction accuracy (using week 1 data to predict week 1 data), we find that the Individual model is the most accurate, with an F1 of 0.526. The overfitting problem is clear when we try to generalize to week 2 data for individuals. We explore the data sparcity issue in more detail below when we consider training size effects. We have started to explore techniques for smoothing the Individual model with the Group or Population models when there is insufficient data, but we have not yet found any advantage over Group model. Higher-order Markov models might be used to improve predictive accuracy. The state-pruning techniques described by Deshpanse and Karypis  REF _Ref87664304 \r \h  \* MERGEFORMAT [7] offer an approach to balancing predictive accuracy and model complexity. Training Size Effects Figure 2 shows the accuracy for topic predictions for Markov model for each group of users (Individual, Group and Population). The data reported here uses week 5 as the test data, and different amounts of training data from combinations of data from weeks 1-4. The predictive accuracy of all the models (Individual, Group and Population) increases as more training data is used. The increases are largest for the Individual and Group models. The Population model improves from 0.379 to 0.385 (1.5%), whereas the Group model improves from 0.381 to 0.409 (7.4%) and the Individual model improves from 0.301 to 0.347 (15.8%). The Group model shows small but consistent advantages. Temporal Effects Figure 3 shows the accuracy for topic predictions for Markov model for each group of users (Individual, Group and Population). The data reported here uses week 5 as the test data, and one week of training data with different time delays between training and testing. The predictive accuracy of all the models (Individual, Group and Population) increases as the period of time between the collection of data used for model construction and the data used for testing decreases. The Population model improves only slightly from 0.379 to 0.381 (less than 1%) as the time gap decreases from 1 month (w1-w5) to 1 week (w4-w5). The Population models are relatively stable over the 5 week period that we examined. Individual and Group models show larger changes; the Group model improves from 0.381 to 0.398 (4.5%) and the Individual model improves from 0.301 to 0.332 (10.4%). The Group model shows small but consistent advantages. We have also examined some finer-grained temporal dynamics. We explored the construction of time-specific Markov models, by developing different models for short-term and long-term topic transitions. We defined a short-term transition as one in which successive URL clicks happened within five minutes of each other; long-term transitions were those that happened with a gap of more than five minutes. Predictive accuracy for the short-term transitions is higher than for the long-term transitions, reflecting the fact that even individuals whose interactions cover a broad range of topics tend to focus on the same topic over the short term. When averaged over all transition times, there are only small changes in overall predictive accuracy. The time-specific Individual Markov models are somewhat more accurate than the general Individual Markov models (0.311 vs. 0.301). We believe there is promise in understanding finer-grained temporal transitions, and will continue to explore models that represent such differences. When analyzing temporal effects, sampling issues need to be considered. In the analyses described above, we fixed the test period to week 5, and built different predictive models for weeks 1-4. Because not all individuals interacted with the system every week, there are somewhat different subsets of individuals represented in the different models. We have also looked at temporal effects by building the models using week 1 data, and evaluating them using data from weeks 1-4. In this analysis, the training models are consistent, but the evaluation set changes. The pattern of results is similar to those shown in Figure 3, although the overall differences are somewhat smaller. We could also have chosen only individuals who were consistently active during the five week period, but this reduces the amount of data that we have for estimating model parameters. CONCLUSIONS AND FUTURE WORK We have reported on the results of a large-scale study of the distribution of topics and topic transitions from web logs. We examined the predictive accuracy of several different probabilistic models. We examined the relative accuracy of marginal and Markov models based on personalized and composite training sets. We considered the influence of temporal proximity. Overall, Group models provide a good balance between predictive accuracy and computational tractability for both marginal and Markov approaches. We are in the process of extending the results with a detailed characterization of the reliability and failure modes of the automated tagging process. For this work, we are validating a subset of the data using human tags for both queries and URLs. We are also exploring alternative techniques for category assignment, e.g., based on the content of the URLs  REF _Ref87322003 \r \h  \* MERGEFORMAT [8], or the content of the queries  REF _Ref87424026 \r \h  \* MERGEFORMAT [1] REF _Ref87322021 \r \h  \* MERGEFORMAT [12] REF _Ref20533752 \r \h  \* MERGEFORMAT [24]. We would like to explore a wider range of techniques for constructing Group models. In the work reported here, we used a simple heuristic of grouping individuals who has the same maximal topic frequency. Richer models that take into account not just the most common topic should improve the predictive accuracy of group models. We are in parallel also working to apply the results to personalizing the search experience, both for the development of qualitative insights about visitations and transitions among topics, and for the application of real-time probabilistic models. We see opportunities to enhance document ranking for individuals based on the topics they tend to search for, and more generally to tailor the spectrum of documents they are provided based on their queries and proximal and more global interaction histories. ACKNOWLEDGMENTS We would like to thank Johnson Apacible, Greg Hullander, Haoyong Zhang, and Robin Wilson for help in the data collection and analysis. REFERENCES Beitzel, S. M., Jensen, E. C., Chowdhury, A., Grossman, D. and Frieder, O. (2004). Hourly analysis of a very large topically categorized web query log. Proceedings of SIGIR 2004, 321-328. Broder, A. (2002). A taxonomy of Web search. SIGIR Forum 36(2). Chakrabarti, S., Dom, B., Agrawal, R. and Raghavan, P. (1998). Scalable feature selection, classification and signature generation for organizing large text databases into hierarchical topic taxonomies. The VLDB Journal, 7, 163-178. Chi, E. H., Rosien, A. S. and Heer, J. (2002). LumberJack: Intelligent discovery and analysis of web user traffic composition. Proceedings of ACM-SIGKDD Workshop on Web Mining for Usage Patterns and User Profiles (WebKDD 2002). Springer Lecture Notes in Computer Science 2703/2003, 1-16. Cover, T. and Thomas, J. (1991). Elements of Information Theory. John Wiley & Sons. Davison, B. (2004). Learning web request patterns. In M. Levene and A. Poulovassilis (Eds.), Web Dynamics: Adapting to Change in Content, Size, Topology and Use. Springer, 435-459. Deshpanse, M. and Karypis, G. (2004). Selective Markov models for predicting web page access. ACM Transactions on Internet Technology, 4(2), 163-184. Dumais, S. T. and Chen, H. (2000). Hierarchical classification of web content. Proceedings of SIGIR00, 256-263. Haveliwala, T. H. (2002). Topic-sensitive PageRank. Proceedings of WWW 2002, 517-526. Horvitz, E. (1998). Continual computation policies for utility-directed prefetching. Proceedings of CIKM 1998, 175-184. Jelenek, F. and Mercer, R. (1985). Probability distribution estimation from sparse data. IBM Technical Disclosure Bulletin, 28, 2591-2594. Kan, M-K. (2004). Web page categorization without the web page. Proceedings of WWW200: Alternate Track Papers and Posters, 262-263. Lau, T. and Horvitz, E. (1999). Patterns of search: Analyzing and modeling web query refinement. Proceedings of the Seventh International Conference on User Modeling, 119-128. Lemel, R. and Moran, S. (2003). Predictive caching and prefetching of query results in search engines. Proceedings of WWW 2003, 19-28. Lewis, D. (1995). Evaluating and optimizing autonomous text classification systems. Proceedings of SIGIR95, 246-254. Open Directory Project,  HYPERLINK "http://www.dmoz.org" http://www.dmoz.org Pitkow, J. and Pirolli, P. (1999). Mining longest repeating subsequences to predict World Wide Web surfing. Proceedings of USUTS: The 2nd USENIX Symposium on Internet Technologies and Systems. Ravindran, D. and Gauch, S. (2004). Exploiting hierarchical relationships in conceptual search. Proceedings of CIKM 2004. Richardson, M. and Domingos, P. (2002). The intelligent surfer: Probabilistic combination of link and content information in PageRank. Proceedings of Advances in NIPS 14. Rose, D. E. and Levinson, D. (2004). Understanding user goals in web search. Proceedings of WWW2004, 13-19. Sarukkai, R. R. (2000). Link prediction and path analysis using Markov chains. Proceedings of WWW9. Schechter, S., Krishnan, M. and Smith, M. D. (1998). Using path profiles to predict http requests. Proceedings of the 7th International World Wide Web Conference, 457-467. Silverstein, C. Henzinger, M., Marais, H. and Moricz, M. (1998). Analysis of a very large AltaVista query log. SRC Technical Note 1998-014, October 26, 1998. Spink, A., Wolfram, D., Jansen, B. J., & Saracevic, T. (2001). Searching the web: The public and their queries. Journal of the American Society for Information Science and Technology, 52(3), 226-234. Sugiyama, K., Hatano, K. and Yoshikawa (2004). Adaptive web search based on user profile constructed without and effort from users. Proceedings of WWW 2004, 675-684. Yang, Y. and Liu, X. (1999). A re-examination of text categorization methods. Proceedings of SIGIR99, 42-49.     PAGE  PAGE 1  Figure 3. Prediction accuracy (F1) for Markov models for different gaps in time between training and testing. Table 1. Sample user logs a. Narrow focus on computer and computer games  b. Broader range of topics, arts, home, business, health, etc.  Table 2. Markov transition probabilities for the Population model. Data from week 1, 6153 users.  Figure 2. Prediction accuracy (F1) for Markov models when using different amounts of training data.  Figure 1. Prediction accuracy (F1) for marginal and Markov models for Individuals, Groups and Population. .01<=[\rs    0 1 B C T U d e z { } ǿǴh2h(LCJ h+{5CJhzoh+{5CJ h{@hzoh{h{hzomH sH h{h+{mH sH h>]mH sH h{h{mH sH  h>]@ hxV@ h#@ h+{@ hzo@h4!h+{h(LhA*S//01= e { | } l ] j  x`gdHxxx x`gd2$a$ ^  g    ( ) K l  8 \ ] j $+,D_Ź|s|jdj[jRh5[hp[CJh5[h4!CJ h=,CJh5[hHCJh5[h'q{CJh5[hxVCJ hxVh+{hxVh+{5CJh5[h!CJh5[h+{CJ h!h+{h!h+{5CJh5[h+{6CJ]h5[h{CJ h{h+{h{h+{5CJ h{#CJh2hXCJh2h(LCJh2hrHCJ_r|S6QY{}.9D}뾷h5[hO'CJ h=uCJhzoh5[B* ]phhzoh5[CJhzoh5[6CJ] hzoh5[hzoh5[B* CJ]phh5[h@CJ h{#CJh5[h4!CJ h@CJh5[hHCJ h=,CJ hp[CJ3"*6, x`gda x`gd,gdWgd8Jx x`gdH4]&#$+D8/0$gd@d84]&#$+D8/0$gd@d84]#$+D8gd@x4]&#$+D8/0$gd@45EFIJL]^vwдݫ~дtdдjqh5[hCfCJUh{#hO'6CJjh5[hCfCJUj{h5[hCfCJUjh5[hWCJUh5[hWCJ ha:CJ h5[CJjh5[hCfCJUjh5[hO'CJUh5[hO'CJh+{h8Jh5[hHCJh5[hpCJ' #$%24QahexyԷǫ限ԇzjzzjgh5[hCfCJUjh5[h|CJUh5[h|CJh5[hfCJh5[h%CJh5[hT_CJ ha:CJ h5[CJjh5[hCfCJUjh5[hWCJUh5[hWCJ heRCJ h{#CJh5[hO'CJjh5[hO'CJU' !-.FGWX[\]캱pɀÀgZQh5[h'q{CJjh5[h'q{CJUh5[h&0CJj]h5[hnCJUjh5[hnCJUh5[hj`)CJh5[hT.CJh5[h^CJh5[h{FCJh5[hnCJh5[hK,CJ ha:CJ h5[CJjh5[hCfCJUjh5[h|CJUh5[h|CJh{#h|6CJ-GYnqcfĻĠ|s|j|aTajh5[hsvCJUh5[hsvCJh5[hg_CJh5[h2OCJh5[h|CJh5[hWCJh5[h{E]CJh5[h}+CJh5[hnCJh5[h4CJh5[hj`)CJh5[h-(CJh5[h&0CJh5[h^CJ ha:CJjh5[h'q{CJU h5[CJjh5[h@CJU./34EFG_`pquvxCGMPRSkl|}ĴĤ͛͒scssjh5[h%ICJUjh5[h%ICJUh5[hQCJh5[h(ZCJh5[h%ICJh5[h9}CJjIh5[hCfCJUjh5[hCfCJUh5[hsvCJh5[h|CJ ha:CJjh5[hsvCJU h5[CJjSh5[hCfCJU&  )*:;?@ABJKcdtuxy|ljh5[h%ICJUh5[hu"CJj+h5[h%ICJUh5[h(ZCJjh5[h%ICJUj5h5[h%ICJUjh5[h%ICJU ha:CJjh5[h%ICJU h5[CJj?h5[h%ICJUh5[h%ICJ* !!!!*!+!/!0!""c"z"}""""""ηΧη΋΂yyoybyjh5[h(UCJUh{#h(U6CJh5[h(UCJh5[h8JCJj h5[h ZCJU ha:CJ h5[CJj! h5[h ZCJUjh5[h;CJUheRh;6CJh5[h;CJh5[hu"CJh5[h/1CJh5[h%ICJjh5[h%ICJU$""""""""""""""""#o#p#~#############_$b$g$y$$$$$$ %% %ĻĻͻͲͲͲ͙͢~uh5[h$CJh5[hJeCJh5[hCJh5[hpCJh5[hACJj h5[h ZCJUh5[hz"CJh5[h~wCJh5[htyCJh5[h(UCJ ha:CJjh5[h(UCJU h5[CJj h5[h ZCJU+ %-%>%A%H%U%`%a%b%n%s%%%%%%%%%Y&_&l&m&u&{&|&}&&&&&&&&& ''D'c'(ֲֻ֩ysmdddh5[hJeCJ ha:CJ h5[CJj h5[h ZCJUjh5[hACJUhm@zh,6CJh5[h,CJh5[huCJh5[h CJh5[ho@CJh5[h%CJh5[h4Y=CJh5[hHCJh5[h%MCJh5[hACJh5[h(UCJ hm@zCJ'())** *(*)*=***+$+*+8+T++++++++++++5,6,7,O,,,-n-r-t---ʸӝӝӝzpfh5[h,6CJhm@zh,6CJh5[h CJh5[h,CJh5[hyCJhyh\h5[h'CJh5[h%CJh5[haCJh5[h\*CJh5[hcCJh5[h}DCJh5[h\CJh5[hQCJh5[h$TiCJh5[hACJh5[hCJ&6,O,/B2I2Y223=4Y4607<78#8M9T9:: ;;xgd"4xgd"4gd xgd"4gd _gdD2xgdD2xgdD2gdy x`gdZ}gda---].^.........../8/E/\///////////A0E0O0Q0R0#1-1111)2>2A2B2W2X2Y2]2k2l2m2ܵܬǬܖ~h5[h\6CJh5[h"46CJh6#hD2h5[h"4CJh5[h~CJh5[hFCJh5[hm@zCJh5[hyCJ hm@zCJh5[h&ACJ heRCJh5[hZ}CJh5[hjCJh5[h\CJh5[hCJ1m2w2|22222222222333 3333v3{33333;4<4=4W4X4Y4]4w4x4|45566=6~666ӮܥϜyyph5[h _CJ hiqCJh5[h CJh5[h 6CJh5[h6#6CJh5[h6#CJh5[hpCJh5[h>]CJh5[h\6CJh5[h,6CJh6#hD2h5[hD2CJh5[hbCJh5[h,CJh5[h"4CJh5[h\CJ+66667 777'7(7)7-7.7/707<7788#88888889L9M9T99999d:u::::: ; ;ͳíé}}ttk}ttkh5[h,CJh5[h0CJh5[h^B(CJh5[h"4CJ h .CJh5[h CJh"4 h _h _h% ha:CJj h2h#]CJUjh2CJU h2CJh5[h%CJh5[h>]CJh5[h-wCJh5[h _CJh5[hjCJ) ; ;;;; ;&;6;Q;R;];j;;;;;;<<<<<<<<<<<<===='=C=D=Q=_=`=h====W>c>ȿȶڶѿڤڛڒډwh5[hTCJh5[h~CJh5[hsCJh5[h CJh5[hJ%CJh5[h1eCJh5[h5CJh5[h CCJh5[h=CJh5[h4CJh5[h%CJh5[hCCJhy hF h* hF h+{hN hF h!,; ;6;H?@@4DG?HKHcKuKMR=UEUbUFY]9``gdq x`gd'xxgdygdfS!gdfS!gdygdz x`gd>)6gd#gd!xgd*gd*c>r>>>>>>>>!?F?G?H?L?\?????????@@@@@+@7@9@J@j@@@@@@@ľĵĬĬߑvpjfbh*hxV hi~CJ h@uCJh5[h}BCJh5[h3CJh5[h/WCJh5[hoCJh5[hsCJh5[hCCJh5[hCJh5[hCJ hm@zCJh5[hCJh5[hSc\CJh5[h CJh5[hSl~CJ hh7CJh5[hTCJh5[hZm CJ&@0A1AIAJAZA[A_A`A%B*BjBnBsBBC(C)6CJh5[hPqCJh5[ht[4CJh5[hCJh5[hxVCJh5[h#6CJh5[hwf6CJh5[h(CJh5[h)CJh5[hwfCJh5[h CJ ha:CJ h5[CJj h5[h CJUjh5[h#CJUh5[h#CJ$FFFFFFFHGRGWGXGcGgGjGkGpGGGGGGGGGGGGGGHHHH!H(H=H?HEHJHKH[HzH}HHHHоЯеооШ|sh5[hBpCJh5[hy69CJh5[h=5CJh5[h6CJh5[hMCJh*hy h*h* hg0CJh5[h CJh5[ht[4CJh5[hCJh5[hzCJ hGCJ h-4CJh5[h)CJh5[h>)6CJh5[hxTCJ-HHHI IIII(I,IHILIRI[I\I`IgIIIIIIII&J(J0J3J8J>JDJJJTJ_JaJJJJJJJJJJJJ.KYK`KaKʸʯʦʯh5[h*yCJh5[hwzCJh5[h WCJh5[h\zCJh5[hBpCJh5[h)CJ hQCJ hCJ h3CJh5[h6CJ hcfCJh5[h>]CJh5[h}BCJ hECJ hCJh5[hy69CJ1aKbKcKuKKKKKKKL&L'L*LCLGLgLjLLLM9MLMMMMMMMMNNNNNNսսճժճա՘սՉwjwZj~ h5[h%ICJUjh5[hqCJUh5[hqCJh5[hoCJh5[hjCJ hiCJh5[hPz^CJh5[h7CJh5[h=CJhwhfS!6CJh5[hqCJh5[h)CJ h CJh5[hfS!CJhfS!hAch5[CJ'jh5[B*CJUmHnHphu#NNNNNNOOOTOOOOOOOOOOOOPPPPPQAQGQHQR RRRR&RQRqRtRRRRR̼޳ުުޡޘ}tkޘh5[hCJh5[h .CJh5[hCJh5[hjCJh5[hY^CJh5[hoCJh5[hc?nCJh5[h)CJh5[h#iCJj h5[h%ICJUh5[hqCJh5[h CJh5[hjCJ ha:CJjh5[hqCJU h5[CJ)RRRRRRRRRRSS S!S%S&S)S7S:S?SSS2T3TTTTTTTTT=UDUEUwgww_Whyh*\hyhF \jjh5[h%ICJUjh5[hCJUh5[h CJh5[h)CJh5[hCJh5[hoCJj h5[h%ICJUh5[hz CJjh5[hz CJU ha:CJ h5[CJjt h5[h%ICJUjh5[h( CJUh5[h( CJ"EUMUaUbUU9VVVVVVVVVVVVWWWWW'W1W7W8W9WRWXX1X?XYXeXXXXXXֹɭ֝ɭ֔yys߂jajh5[hCJh5[h] |CJ hxCJh5[hHCJh5[h:\CJh5[h CJh5[h:CJj`h5[h%ICJU ha:CJ h5[CJjh5[h%ICJUjh5[h CJUh5[h CJh5[h^CJh5[h'CJh+{hJz h3e`h3e`&XXXXYY*YDYEYFYGY\YYYYYYYYYYYY Z ZZ&Z'Z?Z@ZPZQZTZUZĴīwgaw[w ha:CJ h5[CJjh5[h%ICJUjh5[hRCJU hCJh5[hRCJ hi~CJh5[hCJh5[h^CJh5[hKCJ h CJh5[h PCJh5[h]CJh5[hCJ!UZVZnZoZZZZZZZZZZZZZ[ ["[V[\[b[c[d[|[}[[[[[[[[[[[[|o_oo|ojLh5[h%ICJUjh5[hHCJUhm@zhH6CJh5[h CJh5[hm@zCJ hm@zCJh5[hHCJjh5[h%ICJUh5[hRCJjh5[hRCJU ha:CJ h5[CJjVh5[htLCJUh5[hqECJjh5[hqECJU$[[[[[[[\\2\3\K\L\\\]\a\b\\\]<]l]]]]]]___8`9`@`X````ķħַ͕͕͞}tkkg^h5[hs\9CJh3e`h5[hEvOCJh5[h'7 CJ h_>CJ h_CJ h'7 CJ hLCJh5[hRCJh5[h`CJjBh5[hCfCJUjh5[h CJUh5[h CJh5[hHCJ ha:CJjh5[hHCJU h5[CJjh5[hHCJU$``acfhhirkkl.o9svvx {V|S~c~~~$a$gd & Fgd{xgd`xgd3e`gdHgdqrsxgdfS!gdMc3xgd3e``a aqaaaaaaaaaab'b1bRbTbsbtb~bbbb c:c;cccccSdddddee!e#e*eyeúñèßÖ{ririri{ih5[h0CJh5[h*CJh5[h`CJh5[h2CJh5[h=CJh5[hEvOCJh5[hSNCJh5[hRCJh5[hZdCJh5[h#ECJh5[hxCJjhsCJUmHnHuh5[hL.CJh5[hs\9CJh5[hCJh5[hbCJ)yefzfffffffffffffgg g/g8gBgIgigogzg~gggggggg2h3hKhLh\h]h`hahhىysm ha:CJ h5[CJjh5[htLCJUjh5[hFCJU hm@zCJh5[hCJh5[h0zrCJh5[h*CJh5[hMc3CJjhsCJUmHnHuh5[h=CJh5[hFCJh5[h`CJh5[h0CJ h@CJ hrVCJ)hhhhhhhiiiiii@jVjjjjjjjjjjjjjj.k0k2k6k8k9kqkrkkklllll׿׶έέζζέέ΅|sh5[hLaCJh5[hHCJh5[h}BCJjhACJUmHnHu h CJh5[h%nCJh5[hgVCJh5[hg`CJh5[hMc3CJ hCJh5[hqrsCJh5[h0CJhw~ h3e`h5[h5[CJh5[hFCJ hp}CJ*llm'm.mXmbmmmmmmmmnnnnnnnnnn*o-o.o;oDoHo`ohooopRp_pppppqqrrss'sDžy hICJ hG~CJ ht0CJh5[h"LCJ hCJh5[hD CJh5[h,%CJ h#YCJ hCJh5[h}BCJ hACJ hCJ hxTCJh5[h`CJh5[h%nCJh5[hLaCJh5[hHCJh5[h`CJ/'s7s8s9sttu2u:uQuSuWuiupuqutuuuuuuuvvvvvv w9w:wawsw0xxy;yZylyyyyyyyzzz*z+zÿöh5[h$6CJh5[h4CJh5[hCJh5[h$CJh5[h^]CJ h<CJh5[h`CJhQeh3e`h5[hACJ hE7CJ hBmCJ hRCJ hWI+CJ hACJ h5[CJ hCJ hG~CJ0+z,zDzEzUzVzYzZz\zyzzz{zzzzzzzzzzzzzzzzz{{{ { {V|дʴ~nhʇ_h5[h'RCJ h5[CJjh5[h@CJUh5[hNsCJjh5[hNsCJUj.h5[h4CJUjh5[h%ICJUjh5[h$CJUh5[h$CJ ha:CJh5[h`CJj8h5[h4CJUh5[h4CJjh5[h4CJU!V|X|]|i|n|||||}}B}d}p}x}}}}}}~,~O~R~S~c~~~~-6~wpwi h5[h h5[hc h5[hhh5[h6 h5[hh5[hQ6 h5[hQh+{h5[h+{5CJh5[h{5CJ h{h+{h5[hd<CJh5[h>CJh5[hd CJh5[h?CJh5[h)hCJ hm@zCJh5[h`CJh5[hCJ&Ztz<ɈFe̊zgd( gdY$a$gdg_$a$gd7$a$$a$gda&DZӂrǃ΃Rkt˄Hmzʅ˅҅û˴|t|memh5[h>]6 h5[h>]h5[ht;6 h5[ht;hm@zh5[h(L6 h5[h(Lh5[hh(6 h5[hc h5[hh(h5[hT6 h5[hTh5[hp\6h5[h5H6 h5[h5Hh5[hj0J# h5[hj h5[h8Jh5[h8J6h5[ht6 h5[ht%҅d3<͇̇ZhtLjȈ,̽~wog]gV h5[hg_h5[hs+6H*h5[hs+6h5[ht6 h5[hc h5[hTh5[hs+0J j$h5[h@U h5[hs+jh5[hs+U h5[h#h5[hq6 h5[hqh5[h(L6 h5[h(L h5[h7h5[hN6 h5[hN h5[h.. h5[hh(h5[h..6,DEFЉE\eʊ̊1EGoxz؋Ռߌf}ƾƷyrkck\ h5[h( h5[h6 h5[h h5[hh(h5[h46 h5[h4h5[hY6 h5[h h5[hY h5[h!~h5[h5H6H*h5[h5H6 h5[h5Hh5[hD6 h5[hDh5[h..6 h5[h.. hs6 h5[h^ h5[hg_h5[h^6h5[hg_6!؍ !"+,܎ݎЎ|uqfjhMh5[Uhk hihA h2AhAhA hohAjhohAUh@h5[hA*S0JmHnHu h5[0Jjh5[0JUh%I h%I0Jjh%I0JUhEjhEU h^CJ h+{ h5[h+{hsh5[hz 6 h5[hz ( "$a$gdA &`#$gd@h]hgd5[ &`#$gd5[$a$gd<$ & Fh^h`a$܎ގ \]^_$a$gd<$a$gdsgd"2gdMgd5[ [\]^_쵮쪤 h^CJ h+{ hohsjUhohsU hihs h5[hshsj|OhohsUjh*h"2Uh"2h%Ih5[jhCh5[U,&P/ =!8"8#8$% 3 0&P/ =!8"8#8$% P 3 0&P/ =!8"8#8$% P / 0&P/ =!8"8#8$% {DyK  _Ref87664034{DyK  _Ref87664304{DyK  _Ref87664054{DyK  _Ref87664084{DyK  _Ref87664252{DyK  _Ref87664886{DyK  _Ref87664900{DyK  _Ref87665740{DyK  _Ref87694830{DyK  _Ref87665524{DyK  _Ref87665539{DyK  _Ref87665552{DyK  _Ref87424026{DyK  _Ref87515274{DyK  _Ref87515276{DyK  _Ref20533750{DyK  _Ref20533752{DyK  _Ref87515274{DyK  _Ref87515276{DyK  _Ref20533750{DyK  _Ref20533752{DyK  _Ref20533752{DyK  _Ref87516196{DyK  _Ref87424026{DyK  _Ref87698646{DyK  _Ref87261790{DyK  _Ref87423664{DyK  _Ref87423664{DyK  _Ref87498122{DyK  _Ref87498280{DyK  _Ref87501283{DyK  _Ref20533750{DyK  _Ref20533752{DyK  _Ref87424026{DyK  _Ref87516196{DyK  _Ref20533752{DyK  _Ref20533752{DyK  _Ref87424026{DyK  _Ref87516196{DyK  _Ref87664304{DyK  _Ref87322003{DyK  _Ref87424026{DyK  _Ref87322021{DyK  _Ref20533752DyK yK *http://www.dmoz.org/Dd +0  # A2f16#ly-t`!lf16#ly-&?@*:xYMlUiZ"JRTqT'VQ""'I0qʭ R9rBTHRRp C3FʡE5;7]wm˾̼y;4yg`0 qgHD&4ah6bj0Lq} c \YWbadͰ R#p)QF]m$u;DY%/#l CaZXsKU?'Za<>s9go3cTdP'IO8mԾT@oK _y ˔V |I [w"jqrwccioo詮?*e^'29sßjkUZw?),-{ZkdĚYzSΈ ['lU5[&la%¦"@n{7v#Q$1@yī$HGZPq瞔:0-|Cfsvjidlssbq`.ZB0f"1ed+-yc`{??^nnN=Ӵ^F{)_^_o6M8/ Rб/OkP0O?aEjx?Xn{i.Òo.М?X'.=Pw6O7?NmEM No4 wYrGл9S9Saiޘ  %ěG3n4gof6;4tvW=/?/z,@i.; ?q}%Zf.[7>n7kG!NxCsfy}@JX|7֣(n&5^Ԥ w φ~}Í5e>6kX~DwG}`yahuWZyxv_,d36Ҝ+i .I7ƤFp"ws{7fp,3'8EEFם__8Leb-oBt'nz=`u)/6O~#h^pѾ-q4£U}dSz}xG]X4)c&[ABn{-σքwҦS3 }T1oT:6ͱp%wGQ C53nx=sOó2ab mMbɐ?c^rcMu?]  5ʕndn㑡wo̺SP+=U&gt>b~|PoGhBg~&Cʾo2?)lZi<[ֱ5^Y ֫svV;̽"(#h|)(E@=jEXǡ_P2Fcw$"Iֺ[鬽uMN_>#jUo(\j}E@P{Cͩ_B 57ƾRT5W9p p["KsL_n,7߇' 2ՇS+~q&>w"l!'(;p>*;G`", |%Hm ߞxdr@1yjSm] @W訐)@oG'߿[5DERZQ?҉+ǯ!bv<ÛEJ,,.i*Q>Ν)ܑZuj|u[ߋf&U?-tj`X2-İ@Rj[Rt~Q]FEX%ّC=qsb؍4.,pyFdbjh^^+!`)b->^ܜ|{DObk:CtQpJ[xfZR>i#dlh=.b^@q9miۙ.s#4뫯?{QXSΘ?^b׶$d̠oC%ڧֽ-fəj+S Ƃm1xa:;@`"݁I\߿S+Vd jLb3a?hys47êanm'/t$gaXK-v JUE@r0sW+{bY <7F(¿\.ھksǽzcݘ=p)&AYc#](Ʈp1rX@WKe3im9ZmK޽^";r㉜wUL=zbg&9/Ob7w֍IU>F{߻ec=O U)#Dͥ$,.b7%7>e^bmAo~3A-)osLA.QEMs.>=f"|L13PPb?mK79{{JB=TяevUHw 7!A\lJm)v8`oaj[K))Y j_,&ܗj`=рF5@mW X};R+>A Vm-({gAuNPE`"''DƊ:g79]ªEE@X_ݐÊ"T!|||ECa'VWW:ȨHjU&݉'Ϧ:`8_\N'O G`,As&i+ҜUEd1~O{A+;Xa>$y A.$St,tqsdI`Stma~.ߑ{j^݈lWc$#OvŐ+@Щ 5ev")>yǒʾ09nd(L{ wcEg焂xd#}| a{ՕOmL r#hv.\P#~x.Y8o壻x&b}M!ŭ41hqmY09-8U3KVMfxE),|L8ӦqA ܈SbAXۼ1d"Xd' KBŽdgV'.P{B/h)b`TD?i e\ O$z)_%ga;]OzrfBHPQ9]h`}Ű=y3"9qL’) SEScO˰H8>lrt2Mϯg}+wh$gٜZП?/&w)\E9I;'o\n@0GUo˷VQ?ϖ9y|yݫ;(7 mɽ:bė(8YEw7W`v&aL/?Xa\S²x)`5nH̝Uy[?d+ %3J|74;3~]'g6:oN>#߼h'X?d @0g`^* {]o1]#4ܑν(ubbFl-/wa1u>B/|coTAQ HW5osbb->YuG"0No /#I?w緗m?c$^b1X̙s3/٣fT NQ8Rú&70&WNhKf5~'uO/xz|體4}p`"|`*0ᨱ.fG=?dceE/䎞bLWbb.3G0[ yB]/A|-Qw`|K/2~oi_lJ=:0Qk2JEffXw f X##ϕQJ7 g&0B+_C-g? eѷ1aCw,{&'#4d%n2oF@Vұq+ӋaP~pu7;;E`Gk+֫_ź΅-c+mmno5p} 8|N9噁u|&_~NZJ=t("P}EFt9EQª.;"?q^*!/q%/4_'S7B"5_Y4C_{Nnj!$&{= _{ԡDЫ}5l^=hFth9N>M ß}YJv̫GsF_Y{I}X5pLU'U"_t95BGMD7G"r~{f2׺6hEB |): :tcCTM,3fhդ"5:)|D~4Ƿq7_7DGGH7%;!f&iަ4#|va/kPkS56UImlעsD{%dX@򞺝h"&t,}B)\%fM(LBS|7nOI ٔ\l>|ݿ7gz(|~nN/%$mIwCnĨ!#i"067F$̔.^kԄwz\\n,qL;^`e #$bH-cTh4ѿۭpsrrR4Ԍ"P@o&bx1Y'湡䋈2yndz(:qw@"&[ 1D$M& %PA+I:姂w^z8'\k_(b"7'*t5-jM'P%!snP mw4"+"~ _7s8C~5%n}L RrRB[N4|ք7)cNC|[R`;PiT۹㧎 rgqà2ku #ǩb0^3xݲ%1%oC $/[cL'0^D6+[&nl){P]Q~Ax?!_?]F7?\-19վw%/ Rx$R*piݯ|H^j ZM=Ay܈:hȡ4.g$fm۞sr+Kz};$ *&%86H(ʲg9QSb(s얚0"}m.;"41D"7\i~3_! t=Hq>}>*g},n*>vF4 ےs|=1O69}0/5.f#CaݾӑmBUt"8͡4:+b.pL1)ys8t]fŪ]pGoiNh7E`ka+X +=;ٕS" .+"P}7?AE@88Tc!_unD"ka#M+"8_bm@P 54/&Q@\9S.04XVѥe5 &8g:QYFj%%kZX޷Rt5Mk~~N#nqp?j "2 j @7:3PˑuGh^w>;`_&".!RFpKwiՃzsk}ܶ~ LFO, +%Bc%$%;eHK'ldOأ4:m׵mM{;?3i(C6K eB$tыmn&ʸr 2:nB ut#~wH1s;w42t$ca(e4h.s!6Ƿ{<xG`jNF 춎;pO шs&o蜙ZWcM$QZ٧l{~sqÆrNP,Dd%A!nV!rǐɇc[wm?q!;= p7}K;s5x!9cm\1/ Ä nƩ$( cȮ)Un(,Ѥ>"ca9#-Vը'佮hXe{n*$ 6=@Dl!8 1t|r/BߝO4F7_|ݸSa boAx`RC[rt_XPW꾰H#+}2'=D(5<"Bľ 2ZD&g;wQX?a9(eIf %됰bQ$.kuB(ƋhG7VlJ{71R:&4SN?1Gi#Eq:;W3%(Yty`6ơv&g:p } nB6jNX^I0%vMtxb7"id֚lg8v>"Ig;Dv*'Al߽7† BxiRo}ಂ Jg7.dLo)(!}' [PH` xR]{AFr)elF3&[͹ԇl7f,? }|1[Pedraz=S\r|f,n!7|∈/U&U"ukaplrg*](|VIާၘ]UU}ȿN%i\O<jn/Xp-4aA~7KaNɛvLW4Ȭ/?rE}VW +$uQPI$j /suA){ª^uBPBYE@Pv@>S߃ۅH |pcvtwv/5-MFJWvexO]ڎImęGq;rHZ_|D3=ipGVƶ&:QDu9@i g3rSױRuO7l̪w_"~!~:j=pSKtǴ {΁Ee \Y$OI .pQQ,B5)>{"`5 jbU{S7Ԇe!UH@uQ#Ġ=$ޘ%ObC|}wMQg#֚,1BF"Ӳ#6]1yA;6RC',\x3$h$%u1*+pf^7L`Ƣ&"228q}AYr GVUA2}/2_G"R(xB$%͆{ĒE JpDQkDf_%Q||HQ#{ <6^g&ĖX#ShlbKc$966\\zߛ=W4?~1pSv.5OG Q!Jx7S}1 _C A7zӲA\oQNdAizZG[ 2]C7maUt\KsWu#-hlḧ́lNYlIĥ_('ݹyϧ3cie@!;PN|K=D-NƘ,tox Җϵ#oW/œ`~BJП g[b(ejoS1)y2U- >_7? ,~Erv|NlMd'tUƛY/g`$A!vEOZ?'3!KWS-i1-L3.\#;^{JpkKBNL}OM ,q==Nj@xt77OF#W h$ȔwSЄ/< ?YLH{k9{~s1Di!!1me)I)|2(oM_6H&Yl67QnLUmS̫m?(ozr)#QO/zoԒ"WN%W^L(J)c!vJ@8sDA׿`۫fLd䇵xyy{{V pYJ6lCyPveS^_ X1>,3jZw -%L>$lp/%)7_b?'6Oe߻aEO掑W4]A@dl pKЩqSwGp ٹ!3TCT%_/Afd/[eiBl~&;^*8HYn'ȣx EC;>xk“,'ԋ/!H4ڜPɪ: *`Za֟?Lv_N|l//}';\=XKjEȿku( 4F(kr$}FX2 תW~xLE}Uz'F:"]"pL=P:GX({,2-=Tt-p5CiE@PG@C ("L#!{Y#-&:Fi9 u}8oXt5*"'t^s‡8H"H&EY+:3L1~ L.M\_gc/ٝ9E/+ĄNy4̊*/ ;f-GjM CE;wΏera]OD#!튧!~||0Sz ]P9o߀ _A?eӤKn8wgѱ܋Չ-%ۃ7c9 [F~J9{h 7ZL"76 FW\}+@'DtX*e7^f .ޮkHf'X0Ҡ; :&isq6I,TEpcȄ\aab>zތEFGXVͱ`͘ۿ4:ɉ*bG''GթPğ,Ob79Cn_h=ճWNbF&2]Av|rA #N5]x& ~̈AIK}~hoSX7J H`h%\  K> s 3d-ȿPϡ-dֺ30i\={2(KU@㘑1n-H,flЗe=-o_|I|ͯ=JD=ޅ ՟̙(5e*tOaṭU/ȝj9[I k;&I#rK^>ս"DیDoK,XQ`im("*7 fˑ~I59SLky ,uVUUwP_uxdC wEm44`8p|󰍡@wNcpĹ(4UJSz̢ΏhLJw!̮f1D sd7U-Bv27?3*rRjUvq~IkXkL &V%B긟q|LuS C 0t*4ܻJœKi̒9FDy+ha̬zi 9/je ps{F0ײ-\^kM?:l7:1RUHF}(T6w`<F>>83GSP-<<3+~*{ׯ/L:] ԡcW3 pۘRd`>Gv5jiv1k7JhD} $V~lb'38Mu 9kw`P-A_ht sxNy/5u{k| b|wf#nw@Xp-s'Dj\Yzw~R,쮹 P)G@[c?xt}R{\PE!j,]!v8hCiE@PG@C ("j}>''oH#q^@7iδ6NPCF#}27 qƘZt9,jaJ^6|OG6" @s8]=w5Wl{zQ;E%85>0@{3'QZ߃y&23dbͪeCBF` paf?LGEg.7HXK0v)NSrb =} @ s?Zk2bˁOJ3"Nv=|CMHa{UM08\iaT8"0d҆nioJ %Q"mGX]eCIi*b+d;v#GD"*^fP}E=F=h꿮 4C_6NB=0gn4,")- G" lۆttwĽb5LiA(KjMHe_!2j$ 즨ϐ{T\\A}q:?o^U߃:iȒ'MF#םnC>e# w@IpյʆTEl+֪pM%7{g}sgiȬM嫾G R!P."+- Gp 2&h!dJ |m,(dZalVOnL_%: rș%$8||_¸+0蠫 a f^1ܱXw f*쭖0?wb1v J 5|]v&}HD H%%|Ɇdr'\‹_= s+3KHpxϗ0_}{lr4 sgh9Ztd۶!,4nnsoy#8~ 0 /e6dcPr7-#XàJ0wPQõp@aXWPnO`9|ɓcZa=JNWb \>#=EEZZkMa޺ptv{{i;0mY vY ?G_#yoֱCM#w؂rkoj8_M vFP("({{P#"tB@=:""j5.+"P߫҂" N{/(4w-x"a 6cWAEFqY~-.c*kcQg?Z #)OMg!_Ol>Xc$f'-Nia',v~|DT#z8 |:( c("Ft8g"/Vm[ *{@{g!4dv5%Ձ,&OG8%ו v upVl('+QhFi(g 4#[ 1r2& k8wOw@$=ayP)XD2EZf6Jbpl.8uKjWOn2\xEHC?yDgG}P,J7-/?Ղ"0@N)1ޛ'C14Jݿķӓ=Tg4q%_N/̅ OZi a6zD>V񅑢ōOD;$ ld:.hp{Ʊ팶²p0p*ͰJB)?4I ܳɽ)(ߔ>cn"CZloC 2z(CQWdBCx/ fC/U3:DMV@sv=pyel3M6m[4*C [yv=b1$K V8:65t(t/ 6AB^dd>KRòǓo0̈́O{(%=+,PSj.4q:,;vwl9/:Wt@nM(O"\-OMs>%?\%h473yk}B=]ҥ"bTcŃ+" "LUk,=Tt-pUckӊ",XPE`'0 Z:D{8eN֧Ɉ-ѨLZBe@z :_+cߚM]k[5/'KF"H"%pF7r;rpQl.-LpCBvu0]0uj^GLꪧV&YcZf3gӚQS94G>Y1(.3g-WvjaPoJ!p)*lR{w dD7q xOs4QtrrЕo2wn px 70'bB Z뗋aon60%̚-ns#%0ivli(ᄧI/MܓL1Q2kɢ]=EC@y<ፆ^\m, tC}V&tТUXFk/7pp>Ck0|sfssӚ_)+f㧊3ןu=fvÐ9-?_jњԟ~>_oL۶8*| iTҸiJ)faDdT#^F~hHr81հߜ_>nfDWmB6<|@('6 i.vXѩ0sCӳ<|0gN̿#df֊S4c89K"vGH+-h'y-" 0`8p;v(SK6F>\R̿Kr6ȝ)v&J}QU}&FCn "ts:B3Z>j\ճ/3U${o> ۷;GZХk% /$0[趟+`1UQ(6h_D#[J»nh& 5Z T+4Z@%|4|(w=pJ/e|?ckmlN4mGݼbV4 ۷7g4YVP 2B$r2\gAHȏ0ã cP"BG7>'wKJt`$s}C, !O[^꒿.2Đ]-'G=moF"ZnTth3Dub&>%q_ʀŋ+7?=5/9CFH`2}CgË(rg+'LBN8)a /wكH܅~M+3y=E4bHx 饚\)) W/cIOg3+|%z7h t zi6J ')lLo4oc/&bMLI2΄'勦%hޟ)P g61.-oMq-;plߌ2j1J =umfa}3n+0,dց[bř Q -JqUyj~-dXc@mT}1 E@PlPME`U+D@M=v6(h}qE@Pv@^Q,IaQ|=v2m"kmhS[{G$JyxrEv9ajQ6Pc Qg#vK|ĉw# kZV/{|:}#0j7}ZñTK%YakaV^swW|bp̱";QI~ {^%* :6!=AٶnKmv S鄟2r*T|¥;EtA,LVʲxe!wN:6ji,q}-MZÜ&F&_٘C@/&{`\(!7l !ݘGZi5ǎR/{-838D }AP̉cX $pZ`9x_ިxz, h L>ϿWE#>>s-ͭT /^l>пuS!|n&*$o.X.r blYo(`wiP€ *COC}B桊k`9:zR㭖QAR<>W/QBxyy 0RJ}L&hyX'o2v@z5,2Z=ϥ)\~u\Ӱ{mXFN˿<cP߿=3|k1n*0W#Һ,gZ"\@@l~Zի'GY(sp $:ˏ )5ѓ=Vk6hF\v;tuj83/PnֹҾT(2{%1zח7(>߾{e.`M?f!˚ $E(5Vc^>'sKrf ̞YW74L__6P ?DS,e7EOzRV@=n1 %=6Y` ˫cHi3 Mw1!w&BHyuZʋXڡ ˟Y(-iz aw!j}ɎyD5 z\n3vQ&K+(Fti2Rix7\mDH8p㷐Z1悵Bx; ؂&t~n>o 0 ǘ~]dLs}?ޛ7D1-zsՖ"V{܊A("Q4Gm."h``GrW@o^jˊ"7 :(@'TߣUAD*v8!ڴ"(#!E@ ;cHDgRg3ru&'vqf>/b'Zn,.շLW9!6ZIɁJX~Z>k̉{b!ğlBRWІwI !{|R)%E<]NFjbbQk| u!^+sVGfPZ1ӕ!C,au uan%Mєqvμ F8sCˇ.HL{ޗrj0qr%#r/IG@ZYOomo[jÇw@`gC{PÓY2Aeij{dy G?g(n&F_}K%DH)"%/d +nr}- R\ĵ1x`N!JF>k#,~%#k|gK1^ B7vUOM'[v̼9X'|blf},g"y@_?ĒX5|"P]qI]yO pfI\U8}-l 7phvsJj.! X/nma~Fޝ' Mݬw6d)yIb 4zptǔVK*)pts2ql49j^DŃu{o[p~[M.ׂ>6m/rI)|_T?CF@![MbF;ZUPP}f财"({{,20ijt}-XQE@F@PiAE@PVެ1{,2Іj5Gp0{Ȉ=G;sͫm֪AsɑMmef}݃L:<^.h6:8iI, UYVr!egv.9҆jD1(6MPm6vSf́2t]~_4sɑ%;s+},ǁ 5as%pnL}:|Nj SXtwӣm֬o _ {0Մmle`@)$khۦ{$vrjKjQ4"%=\EDhJnsK䈬4<^Lo*sVP1M6'rO|ͥU|qnWWwx A]`؛ elɎjE,)jPd!fF H,Nbnw,9fUAd@e%%a3iSpN!Jߣ$$7V#+S,J.δx qAԬ771qąkSp_Ҡ(vG=""WQɑ|lIrdtSnNE̅VDa3%{^g| }[Tn~ 4&Y?,}cVAGilއ!ݖHZ1iqjMPiugz}d"qy*x[ވcbCO'US^ A1Hdt:Ś*9b1n\j,A؋-H7Rca t;M{ƦI 6 |{&hwH\bBԠ?% & cSCJH$mn #h DU 9c-ߑ?Z]go[z/k_ _jXP=@[~f.("TCg"(o XdDW[5z}-Ϳ~+a g"s;%(K˵;ݹL8Q¢v k |O !ms{sUuYՓB4w(M}H"FO:`SC6( $12+k>?SrM,w7şKxFA~5bd>8Zpܚh&%_obV񓸞!}2{Mݔ1+L~8Z7R[?"Ob 6?:Y4ѦzsU}5- ݆dՎ1PgiuskN"q Fh6dxfS 4ppD &LKR'Cm1j.w %k䚤!f^8+5_Т*ٞ/ƞE O\%}OMQ[t?yut/XɥohfO65d4w pjOtzzCFY~yA- Oq)"E:;(X6;uٮ](pOrXW_HFtwiՁ 5Z'yȰFYMє#Vqsq`|#M[J߉eh( 8&4O8 8y/l7NgY;F(͖OWɮvaSA<;2`-yf,RƢTzqb8Su`&B]%孙_nҲ?ybY|FNZa"=j,I1x~je__IB%II.;鶋9,%bp#!,nM~k22S37 y^Z|1zU\܆gO?7Lcjg/[oW-8͹QWxH]_>"Ѐ@c| >A _>k߀[ 2Q I$a Gvi*,$B6syO}uoEpC'(r֦{coDW_O0M_19Lܔ#`=O|"BüPhGo_݋źܘĞەa7>\ 657]O/##ط{M8Ī9tT*#}-ؘ̆q>)Z֑SE@F@=t("P}EFt9EQª.;"?q^*!0/ww'3N.C]ÛS=*6"ir@͉dӏ |{c5>+>|{nC#s}U+@ qaܶ1E;zO/7S8KYvkNrr{ ى޿~2| EPm_Arlˊ ~6͇kclh6y<`-1}&K=Цg&D wh86s\;hV$)v鲁n:W{EmSD edNxG2}!#z[q4 aD9?[_=:99/D'Y4>г"IIKUac&[ '}pYjD>甀-/A?wM//$^HËCl>bB]AakL_\kt)=g#;wtՂ]Oӽ⻻$FII8c;w`+ / e"X^ro *AHT|D] h)n㕧0Һ7p{i\@nlIѢ@~z}#vĨl4`n<|\|."ih_e{Zj9,>P޳1z3V5S"oW(o35W.@^^E{}jfa SХSyW;gVßd_6vNss}uUo~AigepK"-ǻ%F,uB^kوixQ{Uߣloi홉ZRsbp[l^a|YwLXuUgE#%5!'SIP>X{({,2-=T־VuiE0aRP9޶a<,8pv/χꦎX$ݛh1nmj'~m({@s|7T1=s2I!Obuo |K4eC¸*Oj'Ԃ"l'Rd i,Ճ$GT`[/7]" P,vd©jH|َMoO 9>ʁ"ڵ`kݰ Y%Iېr / d+V78:Hr# .QE`"aٙe);^H=Y2>I$[? eS>.B([C!ZeTpxn6"6W荣mXndzӫ'B}@W jYnYe _lFD!fΘ.͆!;rK܃q ^ E@X {~;%6\&dv(`5* 52Ƀ$E$psqTwNo4}by8FgY+=Kcf'~}j͵!cϫ5pY2[u&'dD}97ͷOrv޿Ub>{Fׇl3jlԀ"k>dv̨ `Gx`svMiE@X ﱚRGE@D@=*Vl9i]Ҋ v8A4CiE@PG@C ("wߕz~wP M DpunVe󴃥t:Mlz#pPk?|-w$ ~`7l;,!PV~ SazfbxeR%g?Rvki? !̀]cqh^jV->2 Iq(k|U?dks;HcfO>J`kaVáNEhݿ~߯.Dw'q߅}hf> mݖ[w_]ք"cT:62y2 ΐo rwsWSh>A/ű!':>Vz 07nzYc13S[ߏNN7r"N-!6n^ tm6&60J<;3W6 'K4r!\:䛹;%HT;{UiXfs U@RG1 M ZɴlűgXW(= #5jCk07{s񂓕KՓvnM/U6Ǧh4h1>\| g 5_6NuBZ[Ĭ) ԟ~>_o"@S֬#G~e7]..Z"~o/iXK ,]ēr@Jh+\ ʮZeuv@rֱhAWŹ'@~vK[`0#C|pyY6'i}j/-)8`8p1HS]pi=ؑz $;?腰gI)98t_qGsOC``oӠsՙْ~]Bsΰ՗ 41~xeԧUOS# 7Y 1Wݙ''ݛT轻=* gX66_-;V3\*08Ey/vp䲼O?.F㚕 h87YYud7zѣIA V>H4Td'5#Ld\ɠnœ9>B7ղW,Z;L(u:c^y?؃cOHc2:Ew`kóK= LNR˩X]G녱Po90iL|Vg\Xp-/6!/W1?/8dX_UYL-z,2z|G> z4GIh-({{ـ;"F@=fC(X0縈jTX UUE`>(@wxb=%sd߀%w󼺯͘Ež F"=̓-ۅ91d[wo6w"  H8\R x|)8AqՖ[z58k'C!ˬd7\,}h'TfDΝ^s;Y3Kc9ղ"F![fؿKMtivF(=~/)tp2\`&HYqGVD;s<.\=Q>3(,1T@blySWw@(BVm.e|G3;l׮A-(GxJ] ~fۿx6j bbc|+҄@XޅTоaąE؀A< 8C_z?xh!a5:{8;pǮZkY5}/I% |Iԝ\4 Wl@8C+՚SAVU)'mmI-&MvAp8y;c)/j-+l1&0-^H/v9V@J[7};#ej@J6CQΉ0PHZ>psԩ2$Wl͢=-tc|?:9Ea*d 8槝n6OBerf ى3,# [*k%0wE1ڄr)(|t) koqiw@a7<8:݉uӹ:[=I0tp'7ۏo:ϙ]DsX`34jV+FÅ鷏n\:AZk7b[6{T_|.ν7S~3Y5~Q IQ2w^{h]q?y4¸ػR`Zy#ƴzLeZK xy[,aPGR*0Oؓ 2a:y80s\{@,`F%BwMq2,f%SK@ҼcPOG1PN΀={x{wO(|v݂kawڻ-8bk)SPzgZ:"( VE@ W*K!j5ԴV"߇QPE@~O!2IIGG>qgo9[oK=XŘ𩲖X,ֱfᱬQ}zo{voabR]-!0'{0:&Jxs܇Hb GhFwW N2)1ZC90N[;0cRglmՁ77w{0eIy$ E0e뎙jtt zΧ^i@c|ȩD#<#WA!q i{erQʤKI/}E-aY{8k37ȷI8=kSIzFDeBG3)F>9F.r¸azbZ05k~b:VeedčUi=h"~)i>lh!]c=Fbǹe8ˡ1ečh%I%$A7Y gP 61 WU1}aL#5#tT}~ ?+MģZ&@`(%BE[EVj5XF7 XCz A3%ki4^$ֆUGp91 PbeZhW-Jh7=^W})ο) TuC0b\MģV&dBc&tnv=Nr1n8=ij$H-  >a?0X˲nNB+c)wCy fsJOƖ5:{pl$ 2 aYEI\=ٗB&Kp#ϠLjwQcVQBA"bG)ufni6Fdc1W{<앂b.@-}qHĘЁ,PuqۏXLMpcTKHoF5S,ɀ=Fƥ,o~u E十** (+-/W{p)wLͶ?. 71 qQ+hlVb *np=|D}񁕕XD[/.x IXHtKA}UP}öjNXD@PXZX/ORE c號]/eE@@~tus+h5#k=]P Xd^:g"jPXZX?eV@@1KE@8<{KvOW/XY^L _Ud} r>8 0982ʼn 3?{s-ͭ{՞"'Lm{}>u12#C&e;Nt<mQmalwށΝک"(*>[vZn'F](s'l/Y7RF Eꀸn7|P}.|FE$d!7c։k5n B9.<8Z\s&2]*5}(6?e(O>U~A-9`fƊ8Un({-0Ͷ  ";H01ȸf;-x)򡱛o2]=@pBIYS +xMd3e(7 |>EwK7X'NM)o l=/&@Oʄ,$m~MVZݖ1 V4r :mćeʈ'Sރ@?eşgI댓xC;[ =b7b(8Ihi%H7ק?ze ԍW_ep4OK؇I Ψ8'S'obҜ=;/6}I+$<#Xx`~n>cz >p~ 1EfWˀEm!0=c7h9Jvkt:@==l:{n$p#}<;[]՛ SKW245%LIj.H;ZlZkAmVt$>m{cQR$pH6咳:U sp<_oAMطS=H||9#wTP^S;rFl֓E4 %jvwL%AgG!b[^o7"Gdd~C_z?x`DX=6Ǚx~9xtf0;lç@R uBD%{XY]Eb-+ sO>\Hؕ˒aIQ[ojx|C{$n*Ii8šu^"y:Nr*]F=bAwpVgq9 G&8ixM &6ԣ;m8o<Д) ^Z?]1B_k>@¬&phi_ZVq`5;y$o-xR| ӄ/XiLJ%gPs|ONϿ~bno&+X0] Movj" 7k0/r'--;^mhɲpm[3^Cg)Uv1R%M777S2 V j$h+M]c盋@KP'N%moMJ3`^.,m(s31(t5)'۠;0[Zhh a^ a蚊&0ME~jVmYVf܇?ƫq`3h1SP|~}AE%S֬dN`1OF rZAAu<Н=ay]2#wՓlbh> }ş`bI^qxPG[@/802nyw᪅Rko'MxםtV-niN1c ~_1>6 ȃ; l@Ǹc[ ;m?qQI_lC .K՟p^V;?9j|6zO{?{ueLjMHnnxn"$WWto )=YRa֕!30}toiYMXAU1]6KÓL\|y\`Z*|` ؅>2w~WS(&x)߿c[~ȯZYKw62@׮b+ؖ/,iԀ{TϤ<7Jyܜ7#qN,cޤ$ZGUgKbY\Vή 4w Y؄ا)l,P>AmJ%AE΅=A£{+$ ~ϤGv° F}uЇϻrxf]tX[$$UrszSr^o©dNK]vVo$N01sN=m7V(YJ9@ 8awX*<9,Z-jr30|;;>v_LX9õD; &ܖyÉB^t]JT~qu8;;|iB$[08xeƺǾ~wT:/B^@$(^i}8b*$=aВJ'|p@bOҒV,XQgpa i:HeL&0%iI ϺgEZZk#uYn44{Ś.+-Pn/{:m| R_r&(iʂka;]Nޔ]z3zܽcO&F@{5TZ^7bޙRv~ 1μԆ{l hmFP!@FWªW}uYPj4^ TE`U4tB'f_E'VQ#e,9xb{I eLTM\KswC 㻣hē:v+ |>CbE P`lG,e) |^v>?f V,|CD1g@;zң7j#09K ?v~ڽi& ̑ $%{ %\o,[w<#% S&Xh?ϓQ{=ēC=2 uuC[t  OH#%;And׸diFa 8e4b/8qt aZn~pz+ `2|zL8$L=zE{ 儓y@ZC^mqerfvt7w_PBE&?—{"4Ŵ"0{9;'f6y4 ynzGK=;|9fh^$Jr%0hEoœ&.%>USGN&h*$ᯀ~J}6ݧ#{0{vhUo>N+';^!58&] Gc!K@ڗvk3rϭkkq )ĔJzKl .l279%]CTt;wЄn6v~Xt[]+re3&?2ɨ;b,jDGȿku9 eOeU, hE`wZ W~x@xvp9}b`{)6",ӟ,TwD;Uh,iTMvNPXdW[5zZk}kӊ",XPE`' HGvLC}I/uCEc¸:7)`5;sĘ|ՈklJיIR6o`u|g\QwK3TJ%*}΄JGII;O]qlYwdv@ᛶm8ڜ0"RZ9*ϊ} 4_qE=OXkU3[2ה;R8Zxh|'Åiw} KzxKػi^WCC:xp㩗yӈKն&[`-5dnG/BV%mO[w,z?ucյf-6׽lG'.6YI28Ak80z=P 7Q@&>'}:,v:E'O0>pq=$h'.ۏ1 ng#"Ps>{a>d M~U2[ɪ|}||kUdWR tx0Q`zpܘ(]QjjĸAnYÔ݂5tc ˃8DmJbL} p7H,+0 uoTDk a Lf(S.c"`u*(i1L=O5oI^{kI.py< OVjwu>APx0s,R5YQ>5q)IlB-mixo7F`=oaW7l\! ÁesN]g/ @Z7\R)S`d%  {mQ^XGpM6q6jUt@kmlGFQJ[4ͯ1]= [ֱ#5Zח>]p-S\Z/*jA|.wab(}G2^?>_l\qط"wUM)P}΀9E@Pv{,2ijt}-XQE@F@PiAE@PV@k|QpLS/+de@NOW osu)T<5Bfti0B\fܨ7%L-22+evzy>M4E,w.0$, AOḣHc!Ro @dוq>\`)C`OK4|ׇ0e+Gfu^nJRȾu2, &`s>cV3+iJZkoǦh- D̺Kdx)?"yDd7 oz86SI塓4v-Yz?[J^Gq&oSy"l-,֪>uلb^G5;A1=3k4c Cس Qw {[&3vė٪i6뼺A$1c .nĢ,Ygdh.u5X]q_Lm3-wszP`'aҲG9{ebTFSz_۪^g8h`:H;,S9#'3~yc\ obK-?!QK9>w:;ɉ1pp"53ykDɯ2JQѣ1 B?飻֝w"܆~-AǓ6RɐV y<sa-΃D1iLnI@Ltv VM9Am2/3#AP&n2s}EdB5Md4^ H\aY%}L9s ?<tfE NH$m&(˓O3m| |"@@`RL%|6dq`ړzۓrIʶqZ9YHk#?:) V"ցٹ020sآ6= .KyG*m;Bhk=a.sǤ1K߃rU`24͙X__X{{OPp7ϳb"(WvNFxnܽ0qe}vt"hbЩJs(7)~(ُ;˩EDھs!B+Z3rZOPz 0=|bH嚾B fPētb:q¦]xv s,=d`Ĩl| bbV˟[ 2w|"t@`z|7{+jm$4Fud_B*0]offKJVnuƨa>`Z2}GоM({ Xx|2d`1,;TA<9yrn2f XtD,뫧-{Zܷ'c/&-O-2[zӸf"tvR,NƃDL=h-|p%M"C?{U 59 m1$!JP=F`b|gW`Q7)YnF<)JH'SF' Hn!-9~Rr^RL j$ECcVS!j_S(~"0dLUr)O=$@m|ʟw9vu_(~aVMYV ZIoՃ7EcLVLCN_Jh1sx2f=map}@Kx`P`8^^'Vdj 3QE@8Tc^unD"j5~N(c"px0Sლ;{@r!/OKvUmȰɯlhYItp?JADϮRMܟ78i%jkd@LIv&cyN*g%{ פ FOt=؎.d!H>b #Saj;vj%ASƍz -}ORӾl{W\ M7e8|}i܉;x9E]OcŒ@5$?H,a1˽~ĮhuL\t1q"/OA]qװG6 %bNx >8e2^WW>5'rϤCXTYA8m-sEL2OlnYX򡇱ELm3>#3kz=ឍ yz?^Qdۗt\Gk$sR:,Z;wf9,IBے8_0*_%8{r\xǂgpP>Hi^~dh۵|fĆOtd m- vl0'G<' ""5ꎽO>߹P)!x}qi`KyQC:dk?õdMMoR 4z(h=׻8dd :!L«kas, ^)[ދ}|~}Az+lS+wFNhmEYByb-ǫXoScNHvrX;=ZX0٬Vb5a@6,v\xSwfTt;ԗl4:ACSضN !.{҅AX-yҐO>__|v΅w)Y꣹\7!3R_}tʉk%X[ѬXҷ5 +/j?E 1h1Y{v<ŷ,ZS 5;5|j7/̋* mvduX '>8뫏~A\: ://\Py+Xauu0g٤ bZݵ* =/It:2 Q$HW*rmn 0"91G=d4!ѳ&6Jq= a8h扲4{M+ؔQe}dgx{aIK 'Gzmo ~\qP: Z1c5Vj]CWӥ`&"t3N ?ʵFnIVooޒcQlPk w׆Uo.tMQo˂:uMӒh|_]4#Shx7DiCDAgE@P:#SCE#IENDB`JDd V!'0  # AbH%ʚb2ՕgnH%ʚb2ՕgPNG  IHDR9] sRGB pHYs+.IDATx^}=r86[\NhThʩv`%.ri .+u9v5J\݃%SNu7@IRY#m("`b kU믿 ?j3&Ul5SEMow(mM("YtڬTqE@Plz.|qgYw8QF*WEt4hGn4[?o>mܶһzFy <EFџkR@4Wa8zCx^ooΒ5BR(c]紱۫7_ͮ;}5v:$uНw7?ݢ,{,ʹTA}quh%f/.>ք@1duy#`:+THxݜ9W'>W=}|n>/cڍI:(ʘ}꺛fϵ3MբuZo [f'`˧0T;c/+$-D1UCNUyg}qw/ L,e3SՇSvb@ΑDTX!`&K*fUhGC " t\~ ,+$}^kޱWlX μ#hUYa4 _Z;Y=814Dfi;lA 8?Y:7^PﻯY&Gn~3 - A[~;<;{iJQ1=[~8sEź7H=lm=vˀ@ idEڜCB4qaOU 0Ê{ijL &w??a̴ˎ77,T!}gpw"]*&vfޙše_3Uɞ_jlfmZ#͉#?ՆZnUk9B{g4T9B?3ȸCw+_(D^%iT {7/)ږδBcA(LU"=+3?OP ^:~$wRY).l3n|~g/380 cX:h]o緟OTx"Dq14MO,C\'Kxc eFY?L%_Qu"uXzԀPKjSÄnU+)V[#M׋oQf64ؚ>p7T;4C/c'~gIר)5o}*FY)'Oof<:W5k͓Ul2ԚQ@,#2)צ7c{{IM("F`OsUu?{66*@PE`(V-z+"p0Ń1VTPBp-v: kUE@&og_e^#9C^zmĪz_شfl݊\_PE`\C(E@PIsJpQѽ;(Q82egas->L'>BJ1Zm"x7Is%,)}(lsJrƀ?겂!5ψ{ 1G 1ACk٧7Q’ԞL(~+ RLc{L5r&(sS, UWBhTIBqDAK C(Y*S5,>7B}oߓ 퉉[Taa/flI+5B7)WA(E<C)R]vo1f-ΞN(&w IfP~w&/u`X2%pl Ɛo}޽\ܞW/R#?cW!(Y=!{d~E;1X>ڈ`BA="ɝ[Tac[@s|cKEok؉U,ƏW%G)*(<^1 YOgcK3̈߾g86/=~ \=к^|H eP쳝pM{ҿBv\ tt (J a(RBc8y(JNr|/w-żr ΐ\A%JyXPQ%Uۛ,&";J)mRސ;piCu{徊 m]I}WyAnef و]qҢUN"c_Z=5mKc20Gipfe `@,Ȣ3 [$kN]ͣ{r.3veorfoEsP{}YPsWnؔ'A=hf=wA3;2Ji|ׁڱtz)yp(6|Jnyaw*s9j6WD6e3QAb~oKiLB,iV̨Yh,%x*)ϊJ5"Ĕv"ZŘs GW0)S\UAÁʊE&o\<-w0:NT-d=N>Nv~%Ù]D3ȑDư*8ܴN`gtlú@e3i/(%n,T|5IcOjvdrW%?щpιf|H$_z*DTIvwjF=eso0ZE\{ I[4améIȅ[fohh4`sքI*⌒PT'b9z;F8 >{ lįa#6'Vɝ w4etE ab6*p눰Y/ J Hé+2yJ'I("Y\Y\¶2qĐ! $2dӰ{L6RV{3{wDq6%zq\wF'sJdW^52N);foϽ26K.1e,Nq@8^w|8Ѝu~f^Q9(LHDFH)*$?˂\:qbUq+y߱S{rEI6NnZ;:P {6|%{CFLVlD&c{*eyDwLRB5qiiьxPꏠXvTW I 4,~azbgh:UAfF|+v0&霧?B )΅^:2U_\&uFK,xLhqLZRDNR6c5v7JS#WNg{&(,E!|Sl}]`9U6Ʈ]Y4Z̞\ώٿXr}ʝ(l!UeTصTI[Cnjv>"|~ըO5 o^RBELuކQ"Dz^5WJ^b @W6WiVPE`yגE@PPE 'C: -FaV%E$ԃy>W=}K&@prkO. ka0+}2K4Y-L4 E@ ŲjFBSMA gC$gKC/M}V 7{껷9PEʵo CcHBdq<>`b5G,JiHZ44doH:&MU<4ml{3\}.-l¾-ȱQls=p3BBh7(z?G;.|yY7rH;L!zDsZ;F򾞽¿ľ8a_H*8!dW_j!J\rDs] kl.SރVմĐZdmbg̜FGXkǞNw$MZ݊hb 7( O@}M0L \h%4H, 視ᯯ> ;|nƢYlBq2/>hykƇ^HVtA'.  OV|֩ޡܑ9Vġ|AD{KIAD]&yOfJZ@ʭ28Q?1tk%Hia{gKD4ǟWX㘨2.\qT) #FO8&zwlO-_UmQ+OO?//i?e{@M㇏ i@IA߅RR_^%t8CYw ťRz(eD:X",'L\;bƛǏ-K)֢gUUjd_}rZ,UM &;*vl%ek$c䧖("5p& 2zwuOkQӤ 'ΑB,0 xРsi}yB܋N8Bf!ymK8aAU_ ̏YhzbnU,)T0pIN?^\~ir-zzn_W)l3ታCwLSkín P3ţ%uoYp{GNNSPJvcDC͏>9LAlrA!qm~,tH8ɂD7?LUEvpdq5n~*kRq8j sW q͆fq] JuY(+14g*ԢgP$ sM`oc~rmh+FN'fEOłdF6!kd|\ץM JHjHfL.kV-\}VLy\ZzrY($@ j\ǔysM,$ ,eP%IEOу싢n*$S)R @0sh?MຼcAKjó:dY'rmOdֲp5̜>nu LɰKcox+x. |k]{E5dvj^ەkV#ޤl!:G>\~!q m*ŹIs$śHI0(Fwt# : /Ub 41G3RA'6f"U?6:EKƒz.:W֛Φ(Q`ςꈥlDKZ,/8?֌; "ޤ!lkXG`),ty} #@/_|ʺUS V nᏯz~uvJqycyV_DZ,r<(PTJ(BkHg^#'C7͚ VeLB$ !ó-w 8Ğ}LWF $ 囨fSXa}6PGj| w,3ݪiL8#Iz^#2~o#:48@ .t#;p2WIj xȂ«slٔ/MUՄE`v's6kppj} ,ݽ݂q_ O:Z; C lkGrzT?[_Ϳ 4;m]3aX'0"'7p/91}Pyi[6LؖYk ߭Hp疝eɹeӣ(?=f I $bu&P@@N: sb4SAbÃTc/ `gIʙ'O K!'̺x>wA0|_͹Jj@Jُ>cIh_ƅqnLupN30vߑ+s]??pR54R ZqM?ÈK!5YFn$1N#l {&Ń7>aIlcͺ>^6V$<ٽЬ&d~ϛT&lU?陝GEA`czMl]M(;FЎR%Ҡ*KʵNiA8wQ`ڹKH9[hkyI{9>n#O w*,d~[b8:sSbZnb}ߙ<< ) /`i.n]ӗGwLS|"`PB<""O2=Mm_g Hz$̻$w߅Wt<+)6X mo%ےߥEJo-=6.@')44Yx,5KsD7P3cC b ͳgfwR:vf>xqqss2O/[ rG\Hw[tD&SD+~?hd/hc4)376:g'F1*UO~uIqN-hz՜AE'ݏMo3Y&!XesuiVp/qMǥbuDmſ徆o[ɭM믱kn 'N7Qئ,(Vۀ"0ϷܲˆGPjPZ4"(Pc.ZWj[TA`}ay3"(G@ϫ6n@U_P@@1g,_$ޤ]e&OҊ|U5>6ᢱ[}YdR+-ݴ:\ʵhLywIMf!y96{=zCn$уq>K&.j:}̏S?34ǜZ䱲g+FNLjغHz6Wq{N9G'-Ϟk1mm3:!YF[ s;(Wr~.>ڮO9,cx9P3d¾͙@D:;hpS2YkX`mCBӆO2xmz월r\hj:Gd0Tl^^mEަU|mdGܡ[EGy~TiB=nkef)P2g$0q06=HpRO̜(+7usOܒT_gNTn 2|!- QlY @ `zric6u3XaW,K/ MOSlZuӷ.G-ݒX,KQf}zv|"x|q}ѡwW8M)֛{)G˜"}8DD\3Ѿ`oC<Ȃ8L>j-nwΈ}h+4$䔘Zh3p8YCۚ#iQ_I1ʵXkX}ŗL%e-{0q0H 'OߟSDpXE=C8Lɏ$t);D?ED;W7/7!_͊J=!P0J(K0A5h-WK^1"lqbeR;]Žc}; %(ע?OO,4n՗}Ҝqa/X?wN)l_%B'v#-I"@˕O `Wf('#%dxY @Sb+Tvp֡gz#eVrDM]~l+A'cs:U׸q; Ԍy%iNRVZ! goQNPqw~n-G:ޅ\.釓˳K{!:a`}bحqRqDŽex_ՌpȮsF.XRVd*(P?+RjK3 c/s~|YxZڷ%j"ëyHbbMLudz08Ly IgHd)#J <-Ŋ;H0k1Dd^Qp #S7{GH1CzElb߰56gPYOE %s;UdKv.|Ɂr-6F Qř\Sڗ$ZE$8J)|8ʀ&qu w(f{bLA'%H 1# o֯UNE`.G( Lh13WFHq-q ۆAQ쉒8v) &IQIɼLʵXDHD=Ye=i}^U_)*T(rC3ݭ%fC _d4S;X@t9[ҭLgU"t^'=̐|"rDOњAQliԒ]D^ɨU(eRZMUo)U[ZS-QDKd L#'"Nh% ?|ue!9D$.G1Q_ą)#B?V*Jȼ R"TTslL>3%jMZ3%CΗ0P2E%[ R65`] f{RD@`&/ j"s *A.wK$^SI;qfUTߘA:")kq-\- j[TA`}a+3"(G@ϫ6n@U_P@@> #2sJk>j O'4p-[`pVX Tnc!.J Et:ԞH #UQw\-!ʵh&w:'z}g >Y#SNelظ Vre1v ;!PQVY6G;%AC4siBS͕Y:M0S;vO×fJy<)dd&V ^ U@zt5:]g' FCȰ%|~sՌsZ>ݲIY!JG%N:&97kn `ht&4e3v A(I͞X1EER%( (=Ffr1$x6㋹;ކ~'G4 ه ;AV=u+hŒHx6Dx&MCHCY+r/o᯸‘{) LК'07bbY6@ Z46*8v7gȆl^gc}g͝s$ys}e%yJ",Tt>ha`/9a~sFC Ҳ5[4٧9 &,D6_?])!D vl?=,MdBS_o=|>7˰b?`(3ұ0Y*&r._@r[X_TJH,(֮#=! #_Cܽ@7K#w`2zNr~*|ظm8|S)g۞B-hb+r6TU6{w)0,H_[S+hZԼ$g k)ReZ^;S)mM\H~Gx~p oCշ?>^]A<_oQ{{E&+$^GmKpy9it0u7ʺiܨ\^t~ >,\q]>w/'p,G[3e!| ܦ+8,5w #~A* D9Y'1*Uo"a-Z(0y=ʝlKxGX+YȒˏ87+cfQ12݇f5zQs&8{P_VN*֦}"\~2f|gv(&rhWOR >W8/;s2Zc/ wF: zv̉ ( dM|)fޞx}}DP}51+,)(j*2Vh<ܟ -g> Lb/y8@Qw v'U)rO"}'@WsS0UhQ"xW0&+aN)feBJ"bmpv'<ΌϵxE$h-IMtoߢ34r}\YL4ˏ!e`D}^2YS)1fʧIb2MaMpS m`zԢj^gbbn_ Vjhd0U %"f [Ab+Q յb?ԥh ^_oo љGeQٖOXMovȷ#oo8M) :z䧁+{{dl>XuLZ[s[o342-Yu0ŏ]̉ZFB(ע E@P"\kjiVZ,z [ _iEE@8z^q"ʵ茜̀i0bʌd1\ *iɅ@?*FRW@d eYRìbRE-^@5֭ !c~;c)m0">M74̅p+ PzpzI.{Cfj @6bPo]0F9ECCH?{!x>0 I2&i0qbʈP9,'SkHGULњ}y>2wYi s-_nqd?%N *9}5 b)gw>lW.fG2SJN \Aa#C z|T1UgM1!=0sD`{qiR!6{ehEi! Qcs 6RԚi߄<*nNV.GG}(zG͋QJ) dڔ44v%l 46v2mM#KrJP%$W_6\E0sz0$@.u75 wN7HOLݍh îJ\L){q +<ܞ{eE 'sr YQZxҰ~t)&&Ɵs$hbifrq+y\} =Z#T}.Oj2LfKt~q gQ盳Ǐ},e:3E 'tKsPE QNX EYF mA[_ 1k_-݉>$W9GO= b {NXYdU*,do9%^x:w)*G$`g% onlSq;R]28OMq xnZ 3w҅2$eTfA_%\;#íadtM~? dֵ I٥%?J/7W.3˹pecO:mK~0bȴXQ<p3`z2:{>@h:+`X)Ws-*L`ɹ˲f U%1!%zVlKL`&.>ժ؝2m9ڗZȊ{eq][m'7cݲ`%1-:{"w. йjYj {&J.-;KvY":ʵ-SE`(ZZ*˩¦WŃgZQE@P6Wm܀"(r-a"O_yh:ȴ#G+!F|7x; g*V)V,\ʾ10SUܔ+r-!CkYF1lq̚w/aG-d͔"YLQWflv(&|D s- :$4y>$e4$ JТ[ S#B]1R2[sO0UՉ]!3R@a*- p\ '-U jSBYʄK{hߐЮgz~Jxrw Q"UȎ2B^Aɘ_:#j$h%r:? sUF<Ϟ'Ҥ7PJ%?y/fVho#7Ȑ)33<ԍbr-.17*"a'32UɶEڷϟ#\Oߜ^A[DؕV BF)ҕ=yc?f-tԙ.g3Ob> CiE!,QzGP;2P)?#Lƨ$$e1cI; c4ͣNgw&9=kq*TERC  Kj$ kF0AZ` mOY}ߩlPLYNnNWL)EiH%&=OX[c Ϋ&=ɲbܪ4B^>w8Fd2߮+#>Ay۴A=Cr-!4wZDhP2E8er4˰!>t yY*m*Q[?fS"UF JpFɓa$j&Q\4IZ(6>qgHAf@27k1Y9TWVrXst-IWv`[ _cOM(" 4v*̠J(" \<E@P֋r-6u%VXN86WL?ӊ*"qjTE@8kQb1G7}46 ‚p9䠼 .jZbw&<}Ydy!IS~ °i)5"8p"~]~ ^^ܿ єb8z$}uQN>]MKǐӮqm!#(Hė5Ʀ €A7as-H .fh T(a+d<^h 8Pd"ZJdg07*:W0$v\=\u.۬O? $wg-d^~~_?/g7WL 5M]^TaHQA rICեM*T&țt c5:U\ΩQ uF *q3 OA@"(9/v΀`b&y՟Յ˭)C`g} 5K^( z8VlTᙼ:˸,<%:CEE4"뤋4J7r-tQmF8j?)y誆6(Hm ; 8]9) DxXiv/ (41l0wnQ 3Wo{|-˴I BY\fĕBe(QE(EaB@5hSUP#R̵f5r+hHp5C9m)Mn(l>=ڜSGaH*Kd 3}"-O2ص SDUQp]ӽʂzKldڈ5l>ͨ0<&J6=!bvd 2 ɯȔb-ynu 5 I_w|i<9"V)؀T>4Yc1ic)K#;YgYMq˸Wid@pef`p^"31$YgR㍩9f!iZܗgᰗv=hE@tI,R!E>wndI }]6sf+i= F(MDЁ3:#LU~ⴅ=/JweSz3mUA^ -4]m*n:r-siVp-Bd{F*Fp0jl/lZ]v{% ͞.:k7:ψ l"TLE@Dv#|T&P&P`E`2ʵ8:ͨ("8ʵ8ĕ֕Z*d@+6Wj*GPE`YkqY|U"(|k t{Fc K)m\̽^\F6(G $qT]*9k`![SXTj0{w DվTɼAeƂ%kZt{F>D !(:'fS^FĎ%w"a+m`(&`4̫RQ#\+#SH-Y-$`+?Iyj̳jF@హa7H0!atAm1 !rv$*D\51!U~{zpM P!N!3s E,"ۄxWW%, (򲉧|^X8f|iz9DMk1xCzj`Y6hQ(, UfilK#|sOD֭žveK*;%v^*Gi\/)}KL+Z2!w5 䚄ICÑQo,o7 4/ݰXLjj ;qi&09yg)O%mV("Y z̾i!!))Tys(0h#b?QGw2s\w7o'9枈!"A缠֝{FRxr)ZL2m eg1Va`iE]V]Ji{yU(1MwMddDX|iW7/TR :[zJLY?:=B\nZ*4,?__#ZZ{CrZ꜎\zwyIƬ!3?︙.k*Ur !G"`RP=O4/7M_0w$hr݋:V0m[yA%m@6iCf3]siZ]B&ILqFUâ? .}]{w{eKLe9Ո3M/k7W8 Ok?PQ7TAI68U٤qv!;C075iP0^7*+Sm/?}B/4>OT?$ ZDi=c-u e\fpYbJ6iR!psE +Pp >JM/fՙpeΫS1Ll?ݴgVZi*a -}9nEpяS nq^~B;uO1~K:s#\(=u<"T5.,(DGiU\1P-*3V#̔*ٻH7y`n97~}+qG۞G'w n1Qʎ?abA2'n/&ZA.W* ZzQ,e+EDl$=tD%?L\qt?{oδr{H7?*4s5,%35 nA`it}B]xvQ+Y&\ Yfk4-ԤKPlc=wʤCFھl͵Ϙyڳ%/SCr8g'mFwaľ -_@lt'NaƩ;%[Z= ph~A为dEr-"c"$) q+UlS#yk;y.j>Rd,cEe`wB?=?roF1h\wMliS(7]^YS(jM"iEƛ[K;핟m&Xd0 RMovfwob}WZVϷT`qD[@~[*m:WMMˉnC5ba^"95ok *DPE`kqX']- jfQ ikqNռ"(C@ϫv("0 Z͵3(ֳ2N3Z);vۊ4&1jx0HnnJv]~ &rq"Z&wBkqעȠXذ1ņ,#"SwSUt ","L*HeiKME,@F,[O<vf>跊y4>ZDOt_PVٳ ~X4ܭ)`D~(EiKzr-V6r-j!-0vwOAI6P'WckΓݘRHTqJ%9JD#*.;l}"fY hcɴgRZ\`\ vzzV!U(9 s-\seף_/G1qRH w8ȕF!e62:D*K"CrD7Wžs-'er@ObC"}bOJ9ˮR>\RKpD]~rG66X%{Dxltpkq2tk?.wn<-\EzV + '0]c@nr8@J?P>pIylP_Җwb ZpVZ_D(^ brJV?ͮnk1L*݂@fb,5/G)%@;vKd sn L+/ewG7'*05OANբL8է0 4Upa%WL a;]˟_QZn/lZ];jNE@P]"kwՎW@ШR8z!bJ X'z9fV!drT%F/:%]L370b D^~.xPhW"8{Q 2Zu I2Q)hWutw Bx@nXx`E/џ,ԺlAęPdբ\|n? +ƗKk߰'ww!4YF<_y4lqCii,( l_5fy&7?SVIDnBHzI_z&s4dX1~ :yo(W4x C8x~>TOLw9Q[Xa ٜm/bRtGxA 6i*d6ע;b'e1J)97K-nJxUJ8G1F9SX.j$Iedplo1`%pr\I:c@="]c4AeL9J\-:ݴE(J:}G״bU *z_شv}`%?{ZD1($4v$4"(PE)"D@GXպWbP¦*햊"(@@ϫa'RPCF>u@mbM)+Icu|n޵mdm_-کq\$|W?FtW=a"#Ib9Am!;q̌L 0kX=5pIhk!* 7хCٲqmC@L1PV5* zCIk-)AJt,)52nþおTZ!U~{ 2lE& lוMG&P!͵H+NczE$6 Da+ɄPUMQvDCH:NKr6|[Bܯ1r22A+/h K*E1C* UJPNϚJ1ZCk8ij׻݅@EPXBz|:lOOu"\UI9hENR[Ho}=1k!mI|@&Ad0r9GDhTDr iCv7/&axl/ T)QNIxg/lg Z)av_.]QE~\}ͦhIy;X6Scx`d3{ּ\ 3C!5(wg;^?I~3HġI G =ID) lsdE^\_tH߅fX~tw<HgDann|чѬLd7]Naa׃ww@ˍ*ޗJي/=m{jBL+sw)]Կު\BRA]ueLg2oߵw86쮨ZO*hfjXvd)ˤ DZy眊ț7p"A@p e䟘S"rlVq" lJ9Z%ԙ2fs\7u{ 7G+էm#eùr[eDxĶDŽE뛬=X|DE7{Ϩ M̟gB>$Nw6;hjlT"E#NÛD0qᨚSbb:dQ A>o4WSShzތ󀜐X)QNr5p2J)= 4娩]RFjTesմfcOSrkOSB827#8Jݸc2-5cm?r̝ᢩHyYrB)#OiBb~NjYRNsĻ+ ώз&K&Yz?f,bD9L^ժDGJ]U겴o+ )|Eh`T"UU3oZsDS'Y؜\AFw "c59sŜ3pmtyH90 c`K#bR)$K[1cpt@`Ýn<-*ߞ^e$өllHHQ)ib(3#8M;݁8kq.ע'@Z41 Ǝ3/'w7jT(gb,qc 4Y"c!úSxe΀bH 1MZCQ:]rD:MGi>;]~9so'r- l\a2 o!TzÏW>(u?\) 3j/gb,qkFL22)ǥY/cT'锦*W3X햴9%DQ"f)%N83(,Erܮ"v_NʕA7 0[ ?.HXg=RvX1RPXf:3U5E zsꟿF4eYJn׃ήSE ZE7GɵJ,re fS T{5]윆q.2lqke*益v:l>+9x}O)}(E7bR2 ͚>0ADz=Z\rf8\U쒆 ~ߜ+vBlp\)cQdӘnqME :28j{) l=,F zzz2ĶuP ͭ9ya鷥R ;EXN⽲ WBB 60/x,&J(8R`R We:uPulb.{_nc<[| 8I)Kd<)Qm_JW=D5MS(SVPioo_l"P"e_#%z"7z8 22IO -QAfX2+6]j-2U%L% rU _ z{Ƙ/㴛ߌpC{`UUP ˱B svlaX2.ٶ,>NUMLC!ʵC(E@P(HKZ*)T" #ikQ"(6mITEPy\v/w,}B//ͫ4@W +hȩlxE:PjZ6XҟQRd +0ըno'!wO\3'~Mc?0ұ Ew癥R^1ɔ$ "Bi1U,}3#4\)ͧ\@\6`yIe-꿑Ok /H[_j.B@ԁ.jF1PD$G;*1+9L ^t_mC6ňTsnŒyP8O> RL$k7*IC/9qsoьTh-I+JU;UE8>M5y6*!$؝)=wD/FRUUZەk F{mXF1޸Ѳp=?EACa:WVs~He cB|PS!K_M$EACrZؐX Jw<38Ij"7TPsB^f{uWp= iSr-Z,v/N|ƵOt8ب43s8 lԅ2bH=^F:DH( keTew~q+ra<uhB7g}~WML1;\X$übw,z_:,|^Eה[gr-bWyv\ŗk2H]zhߦT.?)y8Kl}厭QH(Rlx{Hl\$Lg会I i艏-BQz?-b4Q(gԴQ-Fgs-:?v/~%\{Za HCd DuA; %q$j+'ʼ˜˩B"&piu! G%O0EY9mNHM[q2kQ246~"dV6}r-05 ),Kq$D5G's^ʊ*(]$FH8Jʦ{ɆyG3*ۉk?GVř\1(/hf}r-05 )9V q$y]ɕl_[1 ŲNU 8BBWV"uaI}Q_%9W i_䐜ջ3[$" dmIuߩhZ4wsE Iٲ|7O¿|\i{o4˾LxNn)[YRÉXk֞ҲRUY 4Mpo>=֙HBUZz]{bU'jԝbGt!\u|-&~?Q>\/qtz\02ڰUlAchE@B`}akZ픊"(@@cnN"(r-"Z\VJ^bkq0l/lZ=:~UE`y +"paiT >)FN-v#)hveŠמbsWKG#=V$*FPvas-Jtu]^ _+-&* SzutoQ^]qn5w,/ KvwYT04A#ТMoF Z@,}XIX/*^r5"66Ʈ][wh\=QF%)"0];Aͯ("4ʵ4*_PE`:ʵ89WJ^bmWi [ _ϫۧ&PE@Xz^ 3"(=<"`#{ČnUey~/G$Uٰypdž%,"* R[eUi*WNrK%OHOdݖq(A F%:"?gL;[A%JQfö=~B P?;n1ǩQ0V I2qpՐಆ2" sy88p?^͊XB*OUSѬځeLP8͵)8\U5ta@y{&xB5~JQ9y`xƵZ%"c~爟öΕfy8Vp$*DHRz>}P;xP.gLaXaLKчURa] _U(vz|D%&u,p.r0n)f- bq^х66hf1NTD(s!]g㼁3 "m}ΤS7GRdBGc!$ +ƨu*c)U4ՅHJDJ?^9}>#gOQQRJhYJr1EYX@Shx Q.g%RϢKM - }qXJ+*%v=2xڄ6;_PǷ-i'pEt[F#f{jI7h7" 2NiZ/)u ;~?"q#@>Ž+p!KРGJ%R&0s9Uϕ 3C ML`xroreRG? F/~TBC+մz84s-Y\&Mr3*`yk[QeWӪd)&.8z6 S/ {!ߝ_O)DX"ثi|0 ]~-70پj!2eT<ׅMU%/7+QHI& 7apC+mޛq\vxy5zGjsos7ԌՆ宬1ES+؊x'{K|t?bpF;9%cnĄIps \3G9{zدB=cs'I+*dƖ#C 8 @7IJ+fl\n;i!bB7iYVW֠Y:ڋS^iw/z{Ġaxڹ rAF6`5f B\IomʆfwRbO4G?ڏ=xzi2m;dzCǻGv .7_| ZB3ƍ*Ӱs6C :q:lʵ˯6jۿTC`}akmZ]E@P6ƮݬTqE@PZ<SkEE@ ʵ֕Zb9`z_شz^u0L+(PWEPťKQ CV(mdlkGWkqYEc܂URvp\pԊ@ 6̵Ȩ`){ =Tȱ+>═0xp~)v8G/kB,$d"YvS {5|\$)#ep%B1ʍd!&{Z-?Vjn3+\<,p@ji*V@Tj`XOxwd#ն22[ZP&})g>@s1&Ris=w+RZR.~`T6Ą@ү-0ouwE M $bՕV'< L XR;^G=cHӵz@&: D*<姿vrsz6P Yև;mgQT̛EB~v+ -fR xI-|nIXXLG튓vN1B k\Kp-#ʚZ.nbOLJ;FDui`k\I-- yEG7CT?f?}RBNQN1{mPE \IJjk.7ZTP5Ф{S\w2pPgMvxK\4i9-ɓR?M;lQN?{PQ"BU~kiO_n\2@x~ Vk cax|ۻⵘ: \'-|b>Zm( d;/dўowA*dyioCR/WVrz_'QM M믱kT`X3 E`2v2tq~st{G~tQA@ZSE@Pr-fu%VXN86WL?ӊ*"qjTE@8k>ҍbc@\CO-Nb}ꞂZ- Y je`^c, / @Bεq;1B@&:PYN`0>޼ I!cS LTaY`P|S v:K(¿?AchE`͵G2ysJ_tghώ[!*ZlBg4 +}\u3mQAfʛ ) eSJ{wfT[JG\664>f(k꿧9 ȴ]IN\H 8swRa{;bQMJ,lh[UDeұ4"p2b&#'N3vPS8iΛ8\ T8uI5kw͢=#,?7*)b„O6P=~a4vʌ}YSP쇹t,J)C,4Opw2HwE uw.KFx^DGP(yKxY^6EޚO3'cqN(ac xJ%SQP|.ŗl=)[vh2mGcm#>אCSz/oE)y.g $ZyPITJUKY5>)6< i@i4v\4x#D%u>8>:rAm_Τos΂aսkhe/)S-o#dnU4  3PS>kF*Pey(icۘkqxta%c Wf^Wli(U" [ _c.Ө CfOTk(A@cjKPE@X;ʵv ~"2ʵ֕Zb9`z_شz^u0L+(PWEPE0r`QL_/,hHފŗBPM c!}lqMbV{&\SEQ8x[{.N*Geй/.;[ǟۃ0቉l`"1]%lD/~{|pvGd[iũEJ`ȨZ%wieEǐal 77&UՐv@vqNX9q'-8Ξ8Eُ\ IfP՜SXp;{>uaAN>^_uh1HodTu!9/Ts,@]Wz[iS =wd" $ *#p\Bz  :E96_Ƣ#c8J֟v^QwCqqFJ"s)qϠzqBݬ("?Qv[+ݿPb6v]Zq(iqNZk]t68GtUnbtʓQcچ_I o{^Bkf`G)_huZ*].yGigB,* HkEq68xno3!vqbY}4u-A &#2 i-44LicaUG#P 0<^9yi?EeLy(!6/L'{ 1Xx ĊD "н\CwCRqߖf,"xn捳svl1 ..YqӡŰΆ)NN(ejbr#t~zp#=71p a( );^*`~!.hb|IAiʇ˯WO[+AjO8&zw "K5ӫ 'nɻC00ĘWm u#Xm%=VIɾD9UY֯⸔1|Xf9T<8RzLF%syΑI-՛']&٥88$C(6!;KHYY.6HQNp1֣K~FT/l o/CZ?ww~Jtp֌JN?^\~jD^vԨ$Bd<}SiRZj=&s[rOkt6ey))9.-DjX2pHoAr&/ɗMvV&`sgYǀM]28RX\Pe GMw   }9!uV! LZ5V&#tN|Niv|PHUe Pn$EnHY(g'5izϘ&ݗSBΏIœК7luh . Ad0wbmO iάAlYIg9va8;y~j eBƾ R?ڹ18gj߈<.X*o-*tmMr<~ʵjҟ.ga%Wt5"0 M믱k'Yͤ("s4v!E@PF"\#䊀"(;D@wvoQu%VXN86WL?ӊ*"qjTE@8k\|j'x/y1r-Q駎_|)JO9V҃ԐlccTˉUR7*񛱎-"̇BNp-ZF2rVFhʵؤpoqpDOϜx^{z۬5ʣ(bք[Z9ʄv\Q[_7iUTUSB$P+kE,e ֪lk-qʈ]Z;y΍ET%O̞ѕQj$h A'*HD}hN,tQrrOW56ivUص\ʋMh%+>[(O[Aw0~)KF{>c) y)f`ް&b` ӱ:Ilh08$iADGص)3U:'yXZEqPYaLϟ!v/0' &&kiHfbzCtqO.7VZo2fJ/rf;rۑrrp; oDRr*F %I,3_1w-ڣ uo.Za \ʵ+3>ȧ:&kڃ4B!N?[PCORHe/ʉuψ솘d*Q0yo]=C$ i  r?rM3@_P'F̖ۆe1kq"=-eᡉ2r*"GZC|Y"Gb&DdKIlKRndz;e**2%G^AN.6Hll#7Wr-挺SyQ͹5)mkJ·Q"`Aɳ؊8I(eDT=qLzcUf>@fZ([F`}ak-w}'GK_j5V k7`$UqK~/{-UCuUօr-"(G@VJ^bkq0l/lZ=:~UE`y +"p("Y Cml|ɢ"ۛ^<̵( [1Hײ,(`ʭpYU}]Fz)MoEНl/(\CmÛXJ@0~Vƺ&cB6=b^0p \ng AGG9MT&S}n`UuT(;r"}{R/9->)b`qPl\q֋K#bO^]b131[&ɚ:U7cr-.#zᚆM0"5CGه%ygOǕ%"0]Nqhsld\<^ Xɰϗ/F@K8!s#Te}w &Oe1wRh/LbOPq/Uj\Kp-"Wb.E s-Ҩqv&>ϳ2cYY`8{}uIe!r`Q5qi@lpIJjtcο:[&e7 fE;\ S%-]O2˅ˉ9ݟ;g}L.ROёf123zx`6?4d}ch9Eb3}?r^o6T_4ovup4hgVk* 1eW2%̯t*Uz_شflxom@CĆ&g :T*!U+kZ#sUkDUvhHPd<נCRF5ʵQ("Zl5mXJSBd"/۵>\=.K WWW/l8wU-IWBυ -AaP@am^m-u98kC3ٟ -F[#7 Y(^M&77~bOo-ކ5I\MofZ$҄'{w6&!OфK`V"<:j+5Ƈ:Rqahbt(9"w/.qSݍA, ,Ŵb;/wa;rGS$j'lPVa]O{i[2mC5#p\8t7%; c0%qn钺w4x\#Dd%Xmh\$~;JS:s̢1o>u?;˄F[٥Y\?c@\K,%P$G/(!m4#/:*w@s%%qӊ/̖9mL*G H6 }U1)Ł\ .|)]41;OJ>CpIė^yFu0ŊE~).bP~GeW"҅#3I/[X8Yנ=RHpXЛ0?sWq։&u"nZ5cwR,Zz7^u鎀haKK҂NT*ftm08t/6ܬ% E`OaW1 Km{u|rAnd!ٴfT8kYo͡НjVyu " q /娽rsw'&"c[1[6_ڠN|poY,YY4=Yum3^Lvĩ&1bįWɤV*1{J|NI9'-6ΞT I2fঘXڈ5XB &+C.!"޶P;!ǫx<;`Aoϻ7G7$As-RF_ѿ{qo<6l\T^'ilb`OX>Wb%* ^+;beﲭH`ENrEl/lW_ً$8/ AN!R+526KMzmu $U*mfml/;18'*iq&*-*k&uunoƱH<[Zefh:[{E| hwua}I$8h OP],hص.*G?5"|;0 nkNB&,a{OF/LQ~*Kz$ɼi 断j@مLx_:f,8/<:G2PEEdtgge6@b!\sIYɔ}`,,C/Ѷ7djj-잞QG?2jd#bxtM~DxӍmU>-GXK6 O;tvACUUDJn>WeA1GBLjҊH%> #WS&B-/6fqX(%1A ~jezlUݤeKhOU-Zf:pE3&S:LCfY&sϿ"B29F+G\DNˇ_y = 7/kS2wnp~a)sQ`W*cYJo$^$:\#_|5NؤoqȘH>)ګ$p 5d="KYIiVdh<Ἂ%8q$ar7 k{\daD'ˎRNhf?j~ "\tS%?T~ 2Lw>촟 v7~/RVS[Hn kN=yAjQt_"9 른XRuF]."f5eQLY%ʟ aabn%ݰ2C.mp?AlBl .J|;ZNåp vxbܧ&"nY—oqД5 ƒ2"DsoRNT:Fp=ߢ%>OIP ݛ+)e$oE4쒥}K.JI`^JOU1F#V\5Zps(djQ8,HHFP#n$"ŗuT}[NA]K&y?@˛w>UΫq2`2SUN t"c=%N 7z@tA9uD'?0kvM(͎&Rl?3kWRkZ`vbsD~#3k,iޚ#vYhr? N|-Iۜ:Wl5_zD|"JPGſ4p[j:'4-ymإr@+%tb*oNn6ȌEd0MJYeZmX*+-\&:^}fK_{ ٳƮVf"f~f}Onص܊d0sLtu/jBS5[GuSEPŵպWZ,z [ [iEE@,:Wmt"(r-ѰyDh"EnҘW'kC9d /J\"\hdTMCAڶI/Gd'Cv@f[+7b>6#ifq2ɱ#Ӭ癥R^Qc WytK~'%ΝTR|}|;vM,g+,?#h3evC9.I#elNZҷ']](܃@z9"ъwI3|96kG=AfV[1G}.e>(xWqnYIDS窲 DuE# "nh1We8i Z(6lo^.~i$k2L7Ue1b劈<;͗~t{th TU4b}yJy}MZtmd橖%ZPr-rO 9F<:s^&Ft49~8;7PJE>13pW)z.g\jHI>A0Rd XyY CWvSSٍ/Z-A&^cs!sfҟBGa RA-sX1Ee.uZ(5-'|"d]{ENQ!8>gS?/S :pl]w7i $bAOy(rѪB< G."hgFsP?s<A\Cй`(6ꈟډ/a/ #lo=\sMw ̧!;|dɉ2 nnBB|{gb)j2PGPzBėwb+2DŨ7ؑ@0q\!y(ڀ.`8יؽ`3{`va6B '=4ne-G3,|X1Eu7 NмiBUMh-Eo" ] I$4B 1fq4*ՏO?Ӧ).AO~~aeZ DֆD>uQe qQx? Fz0M[]Ú;I<Ko?P~eſS8%A]|i1 1$㪔A@P0l.s2xbEXyDC3 rWn_T]QcO̺/emHD9 S[^n0 {{\ }eȻsT5Z\kwUvh BokQd3,LO" n^HIݜ}=~};|}UJ ⇋S~-+دvJ?Ǐ?sh^jUZwSH&S`Se67H=ӥfUSr ӡ 9EO_7õSI?ph:ePMڻndn+89O_oLXϡK_QXc >'sXe7-h^^S )vMzZH#C$ Kr-"I;0ԍibJlfJ&+Y !B^$zژT܀AGbZb7Us*"ę } j;\Ng+ 64G#(L%^_ng   %z O'V["#Gz^dϪ\t Fp-(|=fWy-5"*uⴄbvkޔ740:cTWMYˆöp=acݲ'*^IȗVg z ;[J?f?DJH)EyFrF^w,M#yfؼycN-s}2bB8ϠĠ(5-Ժ 7Z,vjJV]YF˩Cko(r}aAwe}X5"P&@m;@@E(oA2QZʵ&k."(1ʵ+y^WؼjinE`4[ "-8jE@P#sՎE@PF#\- hL]h\P>В|$.OOXZTzi .߳Gw*K"g|03yoⶸE>B0d/{[HDv$cV-cH#W&沧A HvwDaLjO/SloY;5 ܴfƀ$pvxlFF"hY3^~Ÿc^S:zmՠ ~5V'zcy<; [e4P'6 cV/.5_ʼn4 X,e$`GpŜ` _` P:Y8DoyQ 8WLjWļz.RHV>uo&@Xóm2B ]Jp6I`̾D3"Dۉd+Jigv*̒iJ)7l( ;,a:. E 2mq3*i-dEhfNK?dQAyO(ع: 010`=\YIyQyz;s/u[?BRk:if ٴf:pE{:I6ZnκB>^0v7X4/KahhD8ēSDHκɍׂY 3K" A8ULD0fiAJ8׵dw~͊D.k9c0-t 䔼Dhe3%"2T\l+&ZEWG#t׏ 80;O8N]2v㻉L,#4{uZΦlxb*TuJ<_Kc>{ sQx_%8Z6(=n'% 9S,9e j?[ݱ` bZ1棰7ucaE 6dlrPQ^h`oVnTξyYķ7K3 d ~ 퍽>-4*x_K@cT.+wo>2AT룥껯YKF|&P=NQAUЇ"_KE1QZĻ~Դ- )%6,X F~/X;x2%DE+=s1,bPPD oƢ,KBukOiTmIׁ '5S ܜ %t_]Gdv 22#[/Kt|N{K6͈=v/ʽ% >annq {Gs*3L}V9"/ n&DNOW ya6tuwuL!1 $@%w/fM q53u` n!~uzSa8āA 9a|mSٰs!ZǍAbO.ǔĞDS}o;a&jU"'(tl,B< ] Eɰ8($LIUk $ELPWxB1" ۱3bƒ .}PZ9?M%+P>^_Yw2fvX9zL[԰#e|Y3jYE`b̏g_oac.ffX/319À(Վ%l7Yg<4E0DфKm#ػ0V1z @BST#6iv:ĻWa7,F׈HҒP[ҢYUtl%gQ Ⲷ:ǂעeH i,?#bT0^d_6kMoWߟWA{GPaٿVVEY?p ׸|'t)\$\WP. JāagUح ^h/_F'$lgeEdݶF3B$V"ՏSF'Lʪ7I ] ˸,-*Y<%PT¼αJʨۄ"jIӠYXJRmd=z:F'.Q=D6B4S8˞Dm09_%ef)ޣ*X 'T$uVPia/Z&XW,T 0|IX2cCcdA~ITBK뗇% o5W, N\hT11C窛 f_!j1<9)%6٢uNa˪,Ô[`{`a1 os=MSbY;jyL1^IZ3@B(FeK)k%-cCZ$|7EVy7зi #XH }eQ?"'eԍ̲*"= ۲CM@ $%#6 ;D{y S2"U. yHǿ*YgC5O@xoպ0`.'`׷=G5ikZ!a:Kq9sW,0aL`*8y;s@ B{TvT21b`lDAʈVem 'r 1hp݄R E ^1^euonnx^ I٠Y &ߘ)ʃ{0AO㇟F5iU_Zl3t{"Ytji'S%gB=&i9jvqBϩ\pd02MWE =)>IAPےZ\ YAr-J`r-?-bYYz\!'0^OU @ې Q&Ɋ<x.LUhoܔi,rgJ/a pʵ89i\Emφ/ ȱ-^g&bvA_;G.H6ٳ=g?BWT!ڌK3tPB@`({p ypus2k%=j F^U$Q-@ڙv.^DpcZ\]ttˬD߂ ./O ` ڇIxmLJ>‹] bJyd w=O2gjb{%LiylE_Q 4W3f4q9ٛc ʵ XmX*Gt6bY_U)3:L¦B6m*L-u-_KSvU;ZQ#`#>[G@ߺ~"lZ\VJ^bkq0l/lZ=:~UE`y +"p(ע3rLE4BOz9 *.gAd18>C`3QGf􌃊'| .r/ ʵhkʵh%HqA3,f=pŷ5.'@:_b(2$4DkqN"`4E Dٔk+4~H#Xy^G_oM-Q)ꐕ΃Ye^"lpNq!{9D}$1!ejf\2v+\ p-a0LmƆ{Qf;h9ޠ@!F· D<W⾓by23ʼ\ޔ #S+`tejƠ .6]lXJ3NͲ64n.r-M6kDB O#TZK&F O nyg&Ü66,^}%|fC$ez捲2HnHX 1xhs6#3L b>ԓXf)2}ziBnkqӸ {-h64B:E דOŭYto >t߭G`ID|H5RULYUV^n+SoC(18DGV7Ho &^RYJ t5Fʵ8(8I,=s*okX^~: $獧as4ϩ?SPolYa>8>:D hQp^b>##BMRϳp?h 2)"E EసųeT&()ESF{>w+Mn"UK24Yj1\HD~<\d\C! ZaT []q6Dr $XdgV'9RLv(ϓ+ŝPI7H3'csHdӿٮJK)%*'fo}񯯏G[F+Gy"54g0hF fS/;MXeަ<' MbFC)Yb;ōٗ%yLAE']F:tYZLg}XIOݻΊ?9%*sht@lZ]ƲgDedQvYd-BA@cjF3|f'Kmn :2ʵx׺+"vkq-Z+yr i`VTP#U7( p\DFI^fEC2O,i/6~7 &-lߠdXJCZ!%V^Bm D7o:%'ޢ5~~;ض^d`e+%UdI2n\c+o!pH\v_t2Ce_SE.x [O)4\ԆTu9p#]W-fyfNJu+ ]"lq >~e"5l<|Vx!Q/>A@p-V,ޤVaj"LiL>NS%:WS-Sw7 ĊMrF"d1%,5-S7d%R]H˓OTDX7 *ѣ] +,'OO?"|S;>_\_t/3Kƴ͉m O?ʎ3ta3">%: <[O΀ \(W e vˎ\1n? 0VSD!k%N~;2SOYJ,_(n\k.2y@=?NewZ^}ʢ%c#4oR'OߟༀOGrO<'xE ݡyU Ddeė>ޱB_M(#,/S$WJB.q1FcQʡmn$Z\2^}! +atwzx5r{g)Ŧӳ0[HqX\3-fwT4VfqoZrJ6{DHpixv/ -!pi?zGXk_ ]v (;<!E L c$Q/ĩ%5C7pbmJ_ OO)l=~};S3J] $pUMWőe(עY'5Z&<>Ƒ*%WF@.!ٳ#YB+xUc{Qʵ(R Zm#S562ƹ[%SJ.Sǿ'|Crh# r`=,-V-Fp2E*ՠC6j[Xg:ܸ1eTnx2J|",̺04NOW`pO3jT FVM $"O5FժqTA1[z83 `Q܁|PYu4|a?Ō׶&IncH󣚆Rb65d^};KyAƛKxDڟ,]T >_|>, V>?$s}erSո>N.ff)A`-(VmZWϠ( (?u1a,W߄*jgE>OBBz!FL#fEfbPR=.t3,@-\ٜM!W߄*h5WHI̙Ko >)עz=b)#cDŨ[|\sn"7xK$FtY  {$GVmzr ⒏g'&3׌hnʵXF3(UZq\nNQO?!>٘^QrS9H{C}rU~A"l;.F\fr-⡼m,㖹ӏhZOi8F(>ܛ=FGCyČ8 jF Pst0Q]}t'8xNѥmyW[K n#'t_עM#oQ}ҪOV:P7A=x Gy%(WM.߉ H .__xZn0}q+˰6*(9{԰{ZsB_FB4\.lTAfLV1VVrfW\+u۳fi5v k vApU" ]p(k{F@l-^PEZ\KX+yr i`VTP#U7( \ɷr-V" }$ e.&PC(r-g6cWr-9k}|")p)~:&8M0/hPҞ&"kkEƈF68wxkCkAZ : 2M"h"U Wr-Vs->>g10ԴZi 'C:?rGDDH62Ώm TCM]b 6V"kQ̠nۂn@R7k1/s*Yiu`ֶgl.HfhǙa ý^Sq^ "ϐ5"k(_&"kQ}7b֐Bp ('x֫5#t%&fDz"%TP FsMs`GhP$(I-0gM mIHZׯxG@E7S1bMg"n?%)_¼ d"k(gpoMqE*עԃDE[vXzcֳ\+NP󎎻>~`˚q4D.C мQ<ܑUk>OI|*0kB̵!s-=I/sY~C(y1|27iVqk%R6j[ Le,\i5vz,X,*]Ʈf~;(c۱ "\oШZ%E@P ʵS֕Zb9Cz^Vj*"py7@P#p\p9嗼 yR RP@C-Gt$FGG=x޼:q^43 0U suv]V$6 _/rr"w$ JL0c} (/Dqk);UoR- w>+JEC`{sĂ*5VyҹW?*?Kb{re9\;iߜB{(f׏0T0yzdhR[5%8F L MXlqnلx67`civoYa82n9Oe26N|sH("3ϟKyb=)B8e *=gqjn;`X4׍7H3hF6AdeEr#ʐI=}o:j7^dޫCi=&ԭg2ȏvnئ2|&6н3Uԑ2pǑU1&&9L,Q6;uZ/}:5%]E7baIzq>EJFp<ۼfT-|0\0U'y /6ۂ~4"yIQRїƑE 5k&U8 4"u*̵Ǧa6S+P}[դmZzNPu\U<jNEݼ9zP %B@DdhAfzMrf>@㭢ҹ+@>4AO䒋vX*fΟDJc&]s.RJ9{eCĀ3 0^7sf&"э6ù8G9\)EGh``v?kpp '墙n&(@(2Q`#4bұb;KTAGokTEۛ܃r'  =339q@D&[t!l=9,ܐ7"`Fؚ2īW@0!TճRA0q%/imZ]ӨmNvc[k A۶lXVļ&ޙ=|/fgN) r;ؙӲLUU@~ª ß9vwg+MsV|}c"v8gyɜ5#)g345'/EX)s ((.TBd C=B5Kp 3w4i哛?Y=}>:bÞWd Xw;עx.FO~#﨩Do()բ 3 e+ K[.<*H,E25\#31mס5oݝѦ05PG`dM 4kס\ɺf8ARoc-e6Ko \0+ ~Oݏ 9sy_ޛ@`}Yc׾6(Vr̞.:sk螂",Ʈ]^8|]jV$7VIXZ\W("Zlb u%V E`[ _ϫFTM("Gjkъ"(U ע -%/Q>-Oi,"^eD\n%CgOgZʕ3]9uY}uVU썕 =dJ5CPиb:ؓ ԈxNlb%eLOiK-"Bw&tPp` '(ycþ$ r\EX?R^{Ny١?{F}j dEQM ҢKfcِ,X׶>Q9 +QǫK^\-C"[!p װ+qnǃsAR{q޴DV}H[ڏ&v/_;uO޸ͽ9sf;;8jEGKyGIfVh =NX;ڡQIC?UẐ08W'^inB@XrJ qZc}58\U7O_jloCtu-;0E.cijomlX)PJ ,-h.ע b&`˷9H Y!$ ybҔJb|/ji \WG_i lQT s} YR( #dO$] ϷQ9 &(˓O!4'U){T⇁Rՠ=)~>mEpzqί.vTG0A#R~pX*) zEa4#!LrX8sʭsDy`ssQ p)WE!vO>HT_%8N;鳔48AI>TS:mN8Эp {o݃%)㕮<שxҹ&˵XiRf:eUdSAU*>#r'J PۓOбMhFFbL;#n$ 0Jv,0UiXyru %D?#֬.姵P_o.<ʼn,[BJ*3|L~k~v"& 1$2ww~;= γL;Bގ"tT+ȵ]e4;P)$W 9TɳG@n\_eY?x !}Y?ܠ2R,+LaCpmU>;*<ǏG'wɬL겇 bABݸN?4 _LJ̊p\ғ'!@7T~r48AGgORlls@.z*bsu~;*3X %#W@gv/H=zZtBF(%4ɜ?ٍ&`6vQɽhBGʵh5k/B8OVz 4ܳg P΢7l+Zl(j搱& Po|3mfX*YB@^)gPLbLڅ aVkqQ _"vwe/1ʭjmCD F$¦n`=7TVQg !T} ۅ Cw{5 l щFZM 7g?? , (M!nBEGbb!C2  "%3:P,Zr-JV3(bfGw^GS")k[ٶ[kZ{rjp:EH9˫5ϐqd*{$ƪ>84WdhV"EJ,%b&+dGƳtUXKYʵ(YAZ??#$'4s-VYH!h.GAI|ӸAe'nc.?<: 8c8݈ QGe5Q0xJƨ,Ы뛭 Mt,a"އw>xZc˵=fo!fukŅҟ~֞W687x]^*; P˘ N k&Fd=}=<jx&vxZgNl;`({{+ |A+I0Ltܾz>WF>8ACܸDW6(mu zE%f'˓k<U@nbAOUb|虭 Yu徥Tʵk6jۿT Y uz_شSj}^~]}T7xa~5y&@SQ"A`[^;)PQ~owڠJYih j{fx8orq*^W&z 967C3ms;|>:RIh *0Ҿrovk,2^teLn1KnBf$lq%7UdV.,>j/so"Xe@b\Y0uFEZmz~B?SYDW)ߟ.E– ^IPm_q3f~y?y8ҰwZM}o@h :T iHqmj/+#Uyy䵓˭;b5E[҉Bx)[Oa0ToجEJJ'>#[Ö[C ^@l&hqR)򡜅V"/xgDf! DWQ)ӈn@L*"bWN0TndZvGqa)[ѿۿįvmKNPV092`W\?Q>]b{W!G =NBfߺ]#2~PR@VGb1v$\ EjZo7L*13|>4HU/*\3 7]K^h/AG2/3tVOKTGW4uE`FnDy ĬF YO仁Ԡr:*d:b˹_ ,̀"EbN%]HQPk ]m?Jޤ {۞1d"2vt׈2KFgMoK"XO|F Au'!R}:-4w7LN9"G[H_ 0E_D0D/I9\}KpVPAug|uV^lj+N1)i(rZ <\KՉ_, ˄>-c1E3A @`Di &/0E-1ޗg $2c ghhNKT]̘,]S.E2H\`Ȑ- ćWU|\]wl'\Ja,B) Z[HcM,n`藨Ib 3( <=: Kݱ&\;Ԙ X3 @/deh(rF)<ļ)e=Zr 9ѶQ2&̀7rcBDQ,4Lt9),iWMrqS0jzWk"zH$ ' Eꦿ;ݼσ!xCkXBu Jԉ >q\ň~dfŁ񰝭$!D4's!\%'g'S4Wc&z&k}QH\T@BYڨԀ_FZػ+נ( $m&k1#1 !2MoDY1:v!e[ B [t*f ygs F *h.!x-bƮFFZٻ(\w4cu ,6pn"R jhd9GX$H nULA|S&d# !\Q8"ϔ_=ۍ&_HTp12nQ[ dC mg`(k:(^fGZػQ$! i9.~OXL9@iO^j5nTc |n~eFdp_:ߕ"NwJ@*hS}E %sjTm8Ry_.Drr^D?㍆c¾!ߢbx 3hQ6vm%Ŀ,iA2cP}W SPMp_!!4IaҬwQ7̩>FO6 #M0~?vȚ8J?5N+!ڂFDDTe$#p> ~l:1cC cj! M KT󵂳T_Ѧ/Ss Xf,+u86 p!9F8-(jÀD.j tevF[/R`dc?|Nth;VpPa_+ ?+Vvş_'D-)&7 Sz# L .G*NuZ 2 Ƞx0M%M%$E@iԹ(V0"$甘^]AC0+ ,pB?ЀHA^ˣu ;ysΡЄP"d!XZ2fe3+T΋ ӥ,zP |!%Sx^b0e ۮqii(tfpb5ÂvTB{k7~?sD'v|;fJ]lUV[D ǟF b _|1 rMrbYߩz֩%%X9ܘ7WR\΅65feey,uRDRk=>Rah)cTfֵAk79S H8@3xa*6k^\=n+m p}KZZd"@2w#:uFj[P/;!Grz(XzffGJZ e gȑZ(lS5g@F%sF}alVP }P?H̶-p* @%;l@sȶ-*|[+7MZFUBY"#?7йبqo'4k$ەc uYx 4 8-;Ghf vggM{mIJ {ZA\ Or`fŸN-nآUqɁM{XԢdp*9PLG-HMd60qNJǮ`L(iۉZ *̀C"`/{ׁ˾vD`%m08-B4qriNVB/ɂӷB c: Jh!@WDq'l _x%WM[|03\rڜ@ 13 ʞ8m*U'b N P҉h b &28Ml⏯l1fHz 1@ c@ŋ'kM &1b e!jR1@ 10'&E 1@ ܆GX4oZiLfy̶V26K* Dmqt) X Gq19ʫHF^_ո+U H 2̼V c0b|P_gZ\?znXyg2n͖~58feq`ŒO*32gʣWȡhao@@[>HB~cAf@Zs[td{,SH 2?(;Sr|^n&j_\ @VWy-8hH%MJmfY+u )s,6fsc+!JK"zT2jzoo$|E?\{|/}0MeYJC;@k,vuU@He);ixãL%.._YTfRd"7~g>^;Iu &989J0!$Bѻ,NZ@7,quV {]] 4`̬6T0Jd̐a->Ќ'!k*ˎc g.i6@nOW Z&NgN?~<w'°?yl:BͨK vv+ݞqDgI!Wxiˁs]~m̏<0iU SD9u Tul*q?ޝkܜ1VMfZH_[w]1f`3hQB ,A9HW37HrFJ<;NXxEKų p/;^;z@f`>. V#}6I=s,ZpBfXnq]\d7Gb7 WPn^gjksݦSoM0em@p7Nvgnd`UQSlQ@/m|'p )-|mzd;H!4S@ qKd@앒gqu;bTx-!Z`Lz#-[&.t܇ R9[88Q̥;Z*P  ( 0 ӳ(S.\N/a WUB!bg-<;Vǀg="@\TXA {2`p#։wz'xQ. 7g7X7Lr1@  Ȩ~yGb b`(jYv b F3-Ul6˲e#]V?1@  pTͿG&9ocq- b/g@Lhؑz'b R(j)e1@  ]IUzeյw:v#Z g9@ j%::Ѫj`ToilQ: DZZٱʝj6[@]]v,1 [ZP(k0d->djLff [Y6X `bsnY d[}Rk9 |j\7ٸhD=rh31xlQIyA!Gv!cwx/o,#i~HJyW L?,'1-Y(iiQ\3[1V 5fAh%T3UZIؒS&878P9[圧봕T:0Wcu8R Ï5|&.Xk1Bᰵ~!*R-8)>\$\%Kʩ&WlXJ7",d9g1ͧ-֣\֊R 8DC<#K _@fW$:ӕ(TUe:˸HS322hVM¸WQ<)tu)@*VK]2%vv+X={Se RK\e{œqx9ufVYAD.8ӯ73.;$u>K8srXBzVr8MUт[˴Rͬu[Ɂ E5K@uZ`aO;s]grrBWZmi/_k9gqǤpcV(݃\cq'v`k3 D,n}xЂ;q&עe++&)󜛔[:vI*-MǕFS@*IhJzV繒X) LK7ZhWNlAVY<{v#:-PW3}k,|EHh`F+(J.5 (L\P֖;i=JU[>}":<M?F%Nwt7JFMS-w>l6S_9N ;]+&!P'3hs髷]1u_A٩tQb-ocaK`3u@؜ =bY|1Lq/#D2"$?jqC,;Øb{%Qd_n[տӚWK:IG-jbNMo#S:{i7,x%g G@ -GWz"(e%5X/Z4Ղ.& HT[:8Bz$/#:%DEQ"hegwl{;a;(ijs38E@bQZj^~Qb b *4B 1@  08-r,1>Ěu (1@  hd?ʘtrX\FX\X\ e !* 1@ 1@ ,E-O1@ @1A1O9PФncey) $RfQrTgˎP*fJRB !>XP:i8Z1*L|'JVŝmrS:m@d %.+>!zJ!|MT 5Yʬg [KƕgaxoGCeLJ=j"&nz8/KKTN O47P "6J@d1jT4ڈ ">SH(!8 +}-@瓝yIٸ n4oQZr "Z : @@+4F[pw3 @#OMJ (s[7oz l=w~DF Zj DKπ>6bn̒ 9 ])"Pqv Sw1%tuz=]5fޒQ xCSY oJXVӯ[ޱ+;j^ֳeteuռ/ޟ.Y-GZ6{}WE-08jAd1WւL=^r]goȪvs6tZe@ -%ʌ"1g"o(T\PnӺjA&.6JChȕiHW1wŽ@lcK{BczxyV{P=JX;̦xqy5C,%唏/zl?=exv>2?r9yguŢ݆,w&[`n%"VZh'UϣocXm6|J/HO.;)ʚ?\/P%%L%btLWBDÿwÍVAE˜]nh4Π٨i ?χ.fDž)}48ՠgq"Zvv+><tw3p3,C$A{|i*sbAnZ"V7_z3~__F?bĬ.1gxlǮ_1֍VeLxO]'Z[ɲPV |?p#/~X//*p#kjF AS"ri(a6+etݹr=Eг/]T۔J!iJij}܉q<]9`#fhu$.$|XO52\a+;rᜅ0x6r}F-boi-#Ȓ]I%Tgx|kmqIO ygmZ/A픑MJ+!zsn0OC  Zo7?г/LCRCCSiCcIN=Ӌxq-tމ!Xk~ex.(QB鍪o "}+zϥ0F7𵥩*뜗):_^3pf8iO6։5"}10xZfQƲM }haȜ6sn\Jm?[|,M(˛e~%~NmfX@lQ3,,هm+_Y& 2v?Fn ~FO6; -Z 볁=%|* [q/ D1-5GfRg"j&Ki >?R?yuA3@D'H_1jih5u@"{]=?0Jx'>n+^@tȦ\Vb!F &"QQQ4}zP'8c&r*mDX)y?kNm3N'r HS.1@W¾ P,;Z xC:v ي<@^Pq7_1qw`h]Kf¨5C4|٦@Dܨ6 0y6"DJ`nXwI,7uo07c5GV>E8b rdRȍ_1@ 1@ gS1@ 1P'TSoleD,S蝔@b`)D{eLIhrX\F`qX1@ 2 %zCTH1@ 8(j&b >$.52/efmg8g`l2o*Y6U8`E`a''7AQJeXZ@Pi:mE0 Ad̰=\߲cAm2[X {zBpS*omj>'0q^SySE'}+3$;TA $jZUdȳ|qnHH=؞ s0łʟƀ`D OiffE UF ܄GyB cejf\=hm8ҺV؀e{3@p3{`(XSaD O~PcF ߇@,D}ʪ-n]q%% @(Z*CsNiu4\8hR: b p&smQ:F:`һ)@ c:f:vv0 7ǘ'6H !ym~UTVvu[w^$yIO5Gt,|GDfWI`n*-t;1Z}mu[S`>z@M>K"@Cg; 3V: pQȗ,-M9|z7tC垻?=cx?q c()n?]DI^y]/Fqwf)G+r7wVp"j]LJ E5Gi⨢&7_Q^}uˡ7UfT1FHL |,٭xdMr?iAeyM!D {k<~,Ҹx)^<:BC~"߿Vs*Ljһ'Md.nJ~`ٟJx-89.SqLm01_^gܽ=wΕr.[_;_ZR,hW1ћ:faW7'|%n>4>ēMv t GT7kՁ _ ߟ9MSbb'[tkс$*ȐC?&JEيz 9WE8 DY]<-WKE$6KM+dx|S;{}4GUuow;qQ% HÁZG,\<٢7e#L.FR1 -Aq/b,ERKܱ:DZǟ(5bͱ+#zЗlFK i~˂D 79*~coğE"Vh,lح|䊡B2}@]I[g?]- ʛobx3K@kPc^[[A:+Y jV2:pfY0[tQ^ 7s峡Yi+wgDoa͚ζ.N56 va8{NL. b Z U+؋4qB;Wq"d9at:|S;yc+d@ %,V6:` 53.:P`OGs+w>[4 F=qv"v*A#G*aw7|í/f:Ab9CU!Վ#5= >Y?>%3chּ][fp9zS yd^͆oI3G.2,@~&|7 %Z0D-wŁ-EU\ZEk?|d()F-)Ip+@\pmiHc?n0C73ufqzDU,j9Xtp&HsJ3, eQ !ʾ-1@ 1p Prb 1@ 1e[ l6bt/g t dże_S5aKs4S2"3.`qbX0G b F0@Q(1@ 1@ ,Zę@V]b M5I19b٢ S V0Ӕ$iL tV`JdBE@4C6E,ylXꔐ,n;"T,Qx+6$0\f.Ѓ2q/ 3ܺ_ 1f)XIt;CTA `7qOKʳwvn9=}R؜x;=7Y}nQ4Kd9{uV !&ۿo x,o:"D18"D -͏'fj9DL9hG-s A*zMF-[3m9Zw>nRQx*Vu8lGD-36Ղ m{aRJ(`<H`"ROP +Z*u@M۽D"=' zpmσ }9{|)Uf \ 0}B:Z"p}yt3E\ +L_Z3Nٷ-,l 6w {/gmw_W~o~quVб.2#lS(l\~ْ5<3EU+T2TuG)3qKfKЈToKP2EaU2WvsZ*|?+c ∸v1CT W_w[̗<%bVnx%m\U;?[2qi&ADBsMi'}UYZyļ VQ2w+>;K;,mHqR3hѾ]m6**TES3gyy)1DiBͿ~g 1r x3扔L: rDG VБŏ[`}{6"z&'E~FQUӋ^2ZZdp=[L櫋"QX42cT7 ȇkMnoj*b=؛_Z0P|+^wD^omU9A+j`1j BtzןGl;jhIb\$"jIhy;)?[9D!s̳K'2ڐN#jʞ|TAr?â ] gHau[!/cM6rxYDВFft5z𔽈 :.oq.Z[tsXaOkvܞeA_š"Td9.ZMT\ RSYmTjҳ e|"F%V^ҕD `k0BE l ԁھvD}3s\l:V@lo Td6+vUa;=&:3_~l rBt6 "0 { gx혥na15vz!֗]v>s.s;W]Ky-]AJ˰ܹ۶<Ɖ(v^2Zu΢߹Hg-&tЀ5Zx 㣖:1h u 0ȵ\ -/#] G$`;[4 # | |qPq[QKՋSnh>?ŕpqd13 Zu1@ 1@ 1<1@ 1@ '4SSl3͆XcAƔ m̰!41p (7{$LrrX\FX @:@ 2 *zCt%$b ZHb b1ZL͹t4PMWh&H=@n l"OeDmԂ1,ocwu"xArQާZ*&l"ïUB+3O.)Y(ַ~zl.cEPv62>zeFib HZGh{0ΩNj Mq?Xp cgb3 p^`4,>p$v DC!gHF:*| hi ;t8q)+8U"J؉M'a5x{8l| 4 Pvk諄Z2dUwӂhp*,%#ɬ^> }Ǽȯ.^fL 0zss&Uec:֥<0X 3-z33 v)xZ`u"JpmEU'A;؇o.zSfTfȨԁa|3~G+`.^vꆗq$,3Hz-(n:Ф9gfQJzk*x,]^;zޯỮ_<ݟ,A,o?cusUdt#I>y/gcWc)nEWLeTpfUJ6*R~(5"d6AFor'ܠlV~V0,s=MW%싧~x|h'й'ۨ\XϞyp#tj{>4~I:5<au3vbU%Ur57vxp"N׺ǝpwǵن`;s]7bم3Tl-zЛY眻'ih` qC>Yї}[@ES}+p.@Nɥ Ѱ=1 ױ3MVg#Q~Eպ}zG BΓ2h MEf 9<%DwlpAS>|ǝ3Q4(!ip-݃\ ^A}9d|n{`MJYPNs>c^7&=3{^wa|w\wx;4Es"fS%y//?.~wfxIXQ_gS>ي'v?&3%`0,- YzAԢԛ$MZO!Ȉ$lǥ$l3BuP- T %uRѮ3Xmy^We#zjKa*?lv9Z%gB-`eJlw\>-Y̶-"]Fe [+lvFuV U^D#Y0`NDoeQa}=o\g[bS\WMv|wUOG2qʀ:s ޗHЬۏM<0f:0tWJ`}`"R+3I`W t&IOE/sہɠgr|i`ćتQvb"CUNo|;_E+6BBB̝^{{ oE泥%qio'{3L.Qr(o\]:!l?:ѳx~>#`xmXwCml nG(- P"d@PGm̰'kok|};6>!(ZO2O9o՜bg`q3\b/g@F-Y]1@ 2@QɊ1@ Ā'Dl6) 1, p #:)PN`d_OIhrX\F   e@FKh%b q P2/*M 1@ K1$Ry^ kEwE o~XS hJ첦0{u1 TAmʼny{਽ DuN=xg}d3K<9 @;?F C9Q7oY_0B+xC+pO>S_pq[{4mf0?Ǖp" ~"P UF-7yY%:r7])M'j3!TT# rpa4B87FK`@9uɨ &WޡƖ 7p?iADs2)"ЇCqJX"(*nƑ俵D Ol }8 tUB +mNtT&Rsf4tE!G\ARs6>ˁZ^&V^%!r߰,O00 `Z[ck65pS/Oǝfv|t5;Qs Bˆ*4%uߏQ0v(nL1&=^Шh.~ +={54LܸiLЏVBlq(;jT"1M<|~S PY\ô<9I^xDrm~| ڑVa|\PS1sB>yr z'UI*FUPTej2s\ٱ1h47{J_w!}wQqwS~> ZPRBt\ izOlv9[H[ /qv;⥃ZVleF&N8nw=[!үRdBe_0kj 爩g`fSn/^esG '`n1p-}CqQW! [t:1W+-o(2#n1Iͭ %J0^ Z厼|~,'] [~dqE)g[dzEP{fr -da _Z\bbrnQpg`NLѵ*d_Rib Y2#bdznUisWYH3,UfQ3̲Ž=a#8lf27n{9%{c~"D쓲!VKE8 9!$e6ݶfq~N>~skݽo'¤5q]"dϕ|;q\rrQe ig: ߌ7?cհ8 7:.g`" 5DĜ:9w33Dns.-m.7*E-zӉs㬺@we|GF*VYΩ`2xSnis¦rQ^`PT:&0\ xb"dZTx(a,vYu: 8϶Q5 N&@dn]̭lp]GmQffXsyr% @,1,ʙ   ,@XP"$b c "b OlI3l%Rb/g6| 2lpTEԷ1@ ,hq,ˀ т#uM 1@ #eYT b dF- <ߤp^ Vz1M^hx-˨sJl H=xi٦Padka`7W@623*UT=H "ȠJ3n0϶~ Ձn(` XI6jq=S3p_:[ OKLmx6E3 *Dpw2=^i#)!(vh7gKm~:pD(--`"hۃ3Y< JMՁ#2v,;ݩk˶5'u#l\ -(~ €z~G- -'8QJVZGZvDR[pNhtJ$uK0tѦZ@[j jN8m ?aqf\W\}p CgZYIt۫.&Ґ >|6'ċ,> :% W#d{Pqg,΢ ;- q? $wQnf9]" )k;F,?J,z5K6k=/Yb*;=QAx9D;ݏ_^Pf~VD"_aͯћݡ̿%Ɨ# if >ws>Qu_q=7,,Ed#up΀*"%LǪ+%,`B_E(e V71D(xs-XhuVdNmՁKT" fS',] -`\&@oyI<2<0T/+{)`ZA^\=nu2: g>YK>7B𵐸_%iT})w=lcb5WBϢ,YlpC gѺb HL|\ƅx0(,h~˂G 79jn<a6H_٦P,Z!f~϶-q XiJVu@TA| m5se-phflDn? /  5HOf`b>G7E6ΈϠ&-3?nZ^[.2n6sGoq@iϥ NaE%JwO{Xۗݱ"IGd XmVR_kߺTe`tDbq3TϚ|O3{Pk =q*|6nqV]}9Qzd.YQ-I8Y"()m31Tf #u%˥  ,{Kc%l*j(zą@L=UTZoq4J8J;"F;'QKy-WM;1@ 1@ -%61@ 1@ '4_kfK͆X3EK">}+nGhWɚ6&51@ {,ˀ G0 #1@ 1@ PB:@ 1@ 23O6YJq <bYcZQ hK~XJqf9QT2ѹgrU:r/4J }NرT݁~z2 %fO30PLEJ l}'w+%Pg(=EF\aC+7tޙ2J;$D?e2>8L qQ ._+F()QKܬV@ }B=!jq[lRFB1/0r5EZEx-ia;BQanX-?(96 -യN30*rqGǒZce|kѴu xLiOQȎmv5C8'>g/a??2l9bt}Ez: й -ED դk\h H̓A˅İ^+X1Y'QN[W|zQ\o12]b>0l,VlIB٦ \Ίwt;<0\fd`sRg&RHp}ڔɡh[tb;2@وR32<,%{ eET JQw)]K4*sRg d{E) :^p%YsD؜ʣ5.,YftwMbJ7:T[j)c6˝VZ~g,tσ-ߜS~tT.Uϱ=cN|y_z?<>u`сf(`V<ޞ1^ K&Ʈ, FؿǫaA:4](]E#e.3>uz==E$TY.nzŭ:˘b2̻c'O;Ec[ܽ='9ܠ>sy;#Mv4^ E#__|\"N8߿o~A%d[Awh(' qB~$)3)GU<*CÁ{H7vs^<7f#l 1<5}UXѾ:kH BaD34{U\5hMxisU{%tSkj|>VGG&|8lE!Z]hW! 2Л,1MIFT@7sn6Az* }]v].lSY%}}aÏE< @B4E#I5! JrzXE"n*( ˤҿޙߘ3޷(et>zf %$xZߌvUB UR^Բ㱧HPe9+sAY (K|XM>=￿ʉ i0\G$r\1L}˅5}zb3/Wrׁ-*^RoVч]>`Hu$d< lG2r)mUV x$!<ôPr3wY=\trW?nw={OS,J8 :0!ʗyT*|yjk]`uq#`jx6nd>L}./&{Cbv+ǝ3w4[4ߜL&{k hDCçFiy=oAcR"$cQ&`QmV^~tBxn򀈛=;J;Tvl(R:|xtVCqsŕBF' {2֕YX/, ̘ҸsA2:0bȪe[zܦV[mEE`9 nr֕KPo^+?/ӶlJ-Ř7HM/ B.i#`ݔ0?ZVݚz:=H1/%WG7HZo7s@Ze VR*Jx5JşEgF*hdaZ6`68WR .N"(P"32P'-7V HW+_ 5mBF_u ԉP^U@gsYZoS6iY8~-t]fގط=3Jݘ]-܄&v ķ<_:QJi#^{XátS依Pw;"@DQMtr.1̯35IW|&)OO6AFsA`v2h u+"!/z"#qsB(ϧ z-o>Қ^|3(xܥ9G#gYi_ gETݓ BK (o~('tсED-V TgK XT#V0IʝڀYϋwP2`u7t꧄{Vt >Hg9Mm>b |wŋP"D1@  3@Q 1@ `[q"B l6b e`xY<)X=Y󗱓;`:I ,.#@ 1\8eu@Fg@ 1@ y(jsD%b bR7쥏nMv%x)-ݴ܏J ST]no}~_uE8V GlU dt`Lu"@k`(g:V0mz@BZm*9vSđv権NB q}M$󎽚_;nֶ'7Ƅh#u䥨;y?U8BJc!@Fh_Aس (S-/|fQJoM^lv$L\X3UVFF3\?%OzZ䒙z$V좥Iv_({Pvd~2rޟnﷻ@D&1Eh:2Y.Ŧ}/"F }LզT|3g4 P`%cmU,5;?/8c!J.E)r˕o;"[LJr(THK vv+B۳d~m5n,%r /DʻH\m6ћs[G9'! DӦq3K<>nVj " VDi ;sW_=|?yue]R3ܔVIWSf!z.QxwH@>>., /&#/6s|ٯG} 8<8t9ȟr22㾇 q4Zy+bmf҃W'}Xn~'旫,Z OS=P0 XŅ:\-շFKW3]J3 7554f8g/y-jv6};R}*Gm*[c0~j/쭣h)>R +.(gP\N@#aFMXNNqsc 'X糼<Ϣb_ճV<@_xP}}aMEQ٦fDb $ኝOV!{# Z?;OKbuձ_)hT|#W^ Yu= )@nl`+\ mov4`ݚ`7V7/Vq}`·n"Yt X_Qh4S6#jɹx`82݉칗S矌"~DDb ݃ԏfE65Lq%󩴳3}PM%Q KFU\95 XDiExD%.wۖm #ʨ`@,\ dy"}UL HQӳeЄ*61ӷn"M2rF ElOeХraT㿤cDN HřS{ ,+QtWG-DGjUuQU-W)JyK%s΢j b``qꀌZ>P"yı!_hbb b |E%#Cbz%1@ '猦H6 ԆZj  > f@ ˾*bL ^^st+  bX|2Z>f#5L 1@ Z>#1@ 1@ `k-r CHT_~M]hx-($fAnz3/, 6J H:SK8Xh~T cҺ %oǤXy˘TXU $  @hWApЛyCbe2pFO{D V9JmL}2I%:!h GRfY>qp "ӞbJ؛w dĞ rj@\"БUU:* g`#4 P.ӁNV7CD3{1@Ar2Bs)8N/{$0CpS`6Ĭ ~WQ+9N( lfczOY/ۭ붓p͂JtX48B7ꬴVYs@yB!ApZZۜ9 A; g-WtfAlŘ_ lWf@x9R #:PT0`nDP =04`"pcrZ<@f6e@B?crrjYT1@ Ѳdr"y!П1@ 1@ hN j'Eg3 Ɉ b 'I3l%Rb e`ѲNſh*rrLNۘ2"1@ N Io b NZN]B b bk- Es=&v0[@樌YBDTлiZ' +BVq\ֹ~ p,۪XK5+b+'[ւ `@fxXSoXHiYoz!͔t{}ڔ2,m E+Xbw4=-(OV1Pbz(!* ̰tj`3",@-LrG@U_v_WBWYfdY@E7{7yχG'lxt#r/Ã\y_>^Yb_Z7 :K8w'°?yXZ_RDro+Xm6!7)[ҹ ^KS5L&ٗviX_o/ŭoЂ\֘65 VUS狀sY|uqs"Ke]]T ?2, i~&Pz@P, }eB@m/|[J{ȸ٣OEe8[`!%DDp (>O`ه:%Гk+6 ]1+(G0{, G2xyG5_>lSmpI]'jب)7057wؿi D#oko]-xuۚMA8E xc,aIAI !Dڊ=)3q`TjP`̯& f@c{'^`|+z¨ ѝvgڲg;i'\w_<=PQO&5J`O,Lणߜ [%RK~4gG=D4LK6;:"ު3*ɥL?aI1.eV+@-5NJ~Lt|JnήSB eMDE`yq\d1}VxB0Q :9 ۖ1`Żʹsgo<ߵ5{c8XʪG[\8 =G:zzvJX7ݟwwp53u? 7Ik^E/蛤Th^Qg"w&N>Jx;<5&l:a넭J a+-HX:aEŠM6EXaNtwaZ>,D=v:Qtj6u 9Éht`7A;U)Ri6]yDÙڛzf:3[] e}j/;b,%c_MDNR!c!hzfG199y5r+CPfxBL wr~,N#Gr`slI#X"Q1ӨƘG,i4'G#Oy؉3܃nYh䲍)M+ F1*bs*?cVc%gZ^ʚu0;LQ#nyN9v9&r,T=K7MoMk1nj4 ?wZOwj|H^m]{Rj+|K_8?Sg+@C-{M6g"Jzv'#\Ye_940v*r,y'k YyDէvb?8NMb퐬~#zM~*T|eZ~[&^Intc+B> ؏ևX_]uM>I5ߏv>Xa1e}Dd 770  # A20y d%HZ U`!y d%HZrR85 0xXoU]nJBQRI(B>Y(*Tm,]LӪHi.9U B ! )zqC2]&M7ơ}oF(,?K4]H1C"ix(FCl| Ǩ7Sf3ny<%AfR +pɥPm6ct2 IZIyi1]\rB?)7A^.U _htǟ I+ŏ۟޸?3UggZm?Ga*zԿxT%CDJ8nNJЯ7{VzrfK(k|kSdimHQ+f "ȖONkUj NB9zs8I5)cLh\$%-L`{v`lu[elc5fe,ϘaNǽ]khߵfR ŅaGiC) n J}m'=<' !:rݸwO97R$-.\5R5m2Rpڲ^oj[rzdT.PJ.v24yKrKNGҔxyB!>(r 9eȏ5;x]̗ո5Wa{Pv`lu[elc5fe,X1]<~OnԼAq\ޱ~Q %M`Q(sB|yFV륹ۙN52*3AfϮD'&' n? G:VYR- ώbGsgٴG6Lm-{&> G.=sL83Tc9C 1x9fB3100Mf*rTqR ـ'YQf8+ȷxf5La̓~{ǼcN﷎H~A9aW,+DF9 NĈ;ι p ^>ŝ]1r{ﯯ _) /ehQ[b2U{FI^GI?F~7{MA}>>J>FXެ,_sn{UGi}=>Sj+%y³9^;N0aŻY}nT}+T1U1h~{KddF&F@F Normal $Pa$CJ_HmH sH tH V@V Heading 1$$ & F(@&a$ 5CJKH8@8 Heading 2  & F@&B@!B Heading 3  & F@& 56CJ8@18 Heading 4  & F@&^@^ Heading 5*$ & F(@&^`a$6CJT@T Heading 6 & F<@&6CJOJQJkHL@L Heading 7 & F<@& OJQJkHP@P Heading 8 & F<@&6OJQJkHT @T Heading 9 & F<@&6CJOJQJkHDA@D Default Paragraph FontViV  Table Normal :V 44 la (k(No List P&P Footnote ReferenceCJH*OJQJkH>O> Author$a$CJOJQJkHPOP Paper-Title $xa$5CJ$OJQJkHFO"F Affiliations$a$ CJOJQJJ2J  Footnote Textp^`pCJOB Bulletp & Fp>Th^`p4 @R4 Footer  !,Ob, E-Mail<Or Abstractc & F@& >T;@ List Number 3p & F8>Th.^8`POP Captions$H#$+Dp/0$a$5CJ<O< References  & FCJ.)@. Page NumberHC@H Body Text Indenth`hRYR  Document Map-D M OJQJ^JB"B Caption$a$5\^JaJtH TB@T Body Text#/H&#$+D@/0$CJ6U@6 Hyperlink >*B*ph4@4 Header ! !FV@!F FollowedHyperlink >*B* ph.X@1. jEmphasis6]HBH  Balloon Text$CJOJQJ^JaJ>R> >)6 Comment Text%$a$PJrh=_1#-0/rh=@1}_.c/01=e{|}l]j   "6$O$'B*I*Y**+=,Y,.0/@FHaKNREUXUZ[`yehl's+zV|҅,_ILMOPQRSTUVWXZ[\]_`abcdefghiklmnopqstuvy6,;`_JNY^jrwx^K   E I ] #x-W[.3FpuR| :?Jtx*/|/(/-/09Z9_9FFFGGGJJJJ K%KLLLNNNNNO&RPRTRURRRRRRcSSSSSS2T\TaT2`\```+rUrYrzrrrrrrrss_X  '!!l12,b$.뜡`qy@1 (  b # # #3"` b -  C"`  / 3 @/. TT TTC"` h 0 # 0C"` b 1  C"` B S  ?aCFQY^c_# 1D-9'%D/ WD074D1D@ 4D _Ref87501283 _Ref87321999 _Ref87424026 _Ref87515274 _Ref87664900 _Ref87423664 _Ref87664034 _Ref87664304 _Ref87322003 _Ref87665740 _Ref87664252 _Ref87698646 _Ref87322021 _Ref87516196 _Ref87664054 _Ref87498122 _Ref87261790 _Ref87664886 _Ref87665524 _Ref87694830 _Ref87515276 _Ref87665539 _Ref87664084 _Ref20533750 _Ref20533752 _Ref87665552 _Ref874982808vvwxzZz{{|t||z}~~<ɀFêz` 8wwxzYz{{|s||y}~~;ȀDd˂y߄`[E\<%]:^DL_ tL`aLbdcTadtTe ]fLEgmLhVij dڒkLlLm 4noTpוq8QrlMs̕t גu֒v4FKT\jss{11CLO  !3z}z}FF`     S[fry}BJNTT   !3}}}}PP`      B*urn:schemas-microsoft-com:office:smarttagscountry-region=*urn:schemas-microsoft-com:office:smarttags PlaceName=*urn:schemas-microsoft-com:office:smarttags PlaceType>*urn:schemas-microsoft-com:office:smarttags PostalCode9*urn:schemas-microsoft-com:office:smarttagsState8*urn:schemas-microsoft-com:office:smarttagstimeV*urn:schemas-microsoft-com:office:smarttagsplacehttp://www.5iantlavalamp.com/8*urn:schemas-microsoft-com:office:smarttagsdate;*urn:schemas-microsoft-com:office:smarttagsaddressh*urn:schemas-microsoft-com:office:smarttagsCity0http://www.5iamas-microsoft-com:office:smarttags:*urn:schemas-microsoft-com:office:smarttagsStreet v 010121998200426296DayHourMinuteMonthYear     178<    U`ek",CImt..44eFuFSS]!]"^-^^^`%`*`1`vvvvvvww4w;wwwwxxx x(xxxxyyyyyzzzz{{${+{|'|H|P|||||~~~~ ɀҀۀYaā́em̂ՂAJ `. ALJJNNSRTRrrww#z.z "+ (+df[`33333333333333333333e}l   J ] $(x-\4FvzR@Jy0m|7$O$B*Y**+=,Y,/./0/;/M1T122 33 3638809`9?@K@cCtCFFGGJ&KLL=MEMMMNO&RRcSSSS2TbTXX2`a```;ccno+rZrzr sSvcvvvwwwwyzGzYz{{R|s|t||||L}z}d~~Oɀ,FЁ̂1zf؅ +f` ` A,!jot}@.@.@..@...@ ....@ .....@ ......@ .......@ ........@1p^`pCJ.hh^h`CJOJQJo([]!joA, pS?A,Aephs?`H O'"L *Sx!~cF Zm } w~ '7 $o@u*GC2FwX\zFp .@d &0( pL.'2)h7"E{D F z 4!fS!"u"z"6#%,%J%^B()j`)p)\**WI+s+2~+!,=,K,x-..gL.T.//2D22"2xu23Mc3"4-4t[4e4=5>)6e^6E7y69<=9s\9a:t;d<4Y=_>l>@&A2A Cz=C#EcGQH%I8JKF:KtL(L%MXN2OEvO PR'ReRA*SuSBTxTxT(UDZUgVrV/WW W(Zp[s[:\Sc\p\]^]#]{E]^Y^Pz^T_]_g_`3e`LaAcOc]j4 .@Hu {#,3'=xVibp d\8QSNz#Y%=@(3MW~yT<}B0| Z'-!y[Eu*wfP!L:Z}:=e;`wzqUw6^{C#i#j:5H!",Jew N|5[ 8$rHOECfq:jt/1&\j?hMz{ t0Q <YuDpJ~}+HIcs>tBpHjG~!9 021eh|E~L(t5g0EJzl.gZn ` U U@Canon i960Ne03:winspoolCanon i960Canon i960 ߁ odLetter BJDM,VT$m,`Oj,`OjVT$m,v`Oj,v,v`OjXXDRAFTSample 1'dVT$mVT$m@ VT$mCanon i960 ߁ odLetter)15Canon i960 ߁ odLetter BJDM,VT$m,`Oj,`OjVT$m,v`Oj,v,v`OjXXDRAFTSample 1'dVT$mVT$m@ VT$mCanon i960 ߁ odLetter)15p_P@UnknownGz Times New Roman5Symbol3& z Arial;& z Helvetica3z Times5& zaTahoma5 Miriam;SimSun[SO"hKFqDqD!4d2qHX?gL.2Proceedings Template - WORDEnd User Computing Services Eric Horvitz   Oh+'0  $0 P \ h tProceedings Template - WORDEnd User Computing Services Normal.dotEric Horvitz2Microsoft Office Word@F#@D5@6@6q՜.+,D՜.+,H hp|  ACMDG Proceedings Template - WORD Title 8@ _PID_HLINKSAdOShttp://www.dmoz.org/  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz|}~      !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~Root Entry F`;Data {e[1Table\WordDocumentSummaryInformation(DocumentSummaryInformation8CompObjq  FMicrosoft Office Word Document MSWordDocWord.Document.89q