{"id":238042,"date":"2016-03-01T00:00:00","date_gmt":"2016-03-01T08:00:00","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/msr-research-item\/a-diversity-promoting-objective-function-for-neural-conversation-models\/"},"modified":"2018-10-16T19:57:00","modified_gmt":"2018-10-17T02:57:00","slug":"a-diversity-promoting-objective-function-for-neural-conversation-models","status":"publish","type":"msr-research-item","link":"https:\/\/www.microsoft.com\/en-us\/research\/publication\/a-diversity-promoting-objective-function-for-neural-conversation-models\/","title":{"rendered":"A Diversity-Promoting Objective Function for Neural Conversation Models"},"content":{"rendered":"<div class=\"asset-content\">\n<p>Sequence-to-sequence neural network models for generation of conversational responses tend to generate safe, commonplace responses (e.g., <em>I don&#8217;t know<\/em>) regardless of the input. We suggest that the traditional objective function, i.e., the likelihood of output (response) given input (message) is unsuited to response generation tasks. Instead we propose using Maximum Mutual Information (MMI) as the objective function in neural models. Experimental results demonstrate that the proposed MMI models produce more diverse, interesting, and appropriate responses, yielding substantive gains in BLEU scores on two conversational datasets and in human evaluations.<\/p>\n<\/div>\n<p><!-- .asset-content --><\/p>\n<p>[<a href=\"https:\/\/www.microsoft.com\/en-us\/download\/details.aspx?id=52375\">Dev and test set<\/a>]<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Sequence-to-sequence neural network models for generation of conversational responses tend to generate safe, commonplace responses (e.g., I don&#8217;t know) regardless of the input. We suggest that the traditional objective function, i.e., the likelihood of output (response) given input (message) is unsuited to response generation tasks. Instead we propose using Maximum Mutual Information (MMI) as the [&hellip;]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":null,"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"Proc. of NAACL-HLT","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"Proc. of NAACL-HLT","msr_doi":"","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"Jiwei Li","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_original_fields_of_study":"","msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2016-03-01","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"https:\/\/arxiv.org\/abs\/1510.03055","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":2016,"msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_match_confidence":0,"msr_microsoftintellectualproperty":true,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[13556,13545],"msr-publication-type":[193716],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-238042","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-human-language-technologies","msr-locale-en_us"],"msr_publishername":"","msr_edition":"Proc. of NAACL-HLT","msr_affiliation":"","msr_published_date":"2016-03-01","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"238392","msr_publicationurl":"https:\/\/arxiv.org\/abs\/1510.03055","msr_doi":"","msr_publication_uploader":[{"type":"file","title":"1510.03055v2%5B1%5D.pdf","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/06\/1510.03055v25B15D-2.pdf","id":238392,"label_id":0},{"type":"url","title":"https:\/\/arxiv.org\/abs\/1510.03055","viewUrl":false,"id":false,"label_id":0}],"msr_related_uploader":"","msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[{"id":0,"url":"https:\/\/arxiv.org\/abs\/1510.03055"},{"id":238392,"url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/06\/1510.03055v25B15D-2.pdf"}],"msr-author-ordering":[{"type":"text","value":"Jiwei Li","user_id":0,"rest_url":false},{"type":"user_nicename","value":"mgalley","user_id":32887,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=mgalley"},{"type":"user_nicename","value":"chrisbkt","user_id":31423,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=chrisbkt"},{"type":"user_nicename","value":"jfgao","user_id":32246,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=jfgao"},{"type":"user_nicename","value":"billdol","user_id":31229,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=billdol"}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[144736],"msr_project":[604608,393245,171447,171065],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":604608,"post_title":"Microsoft Icecaps","post_name":"microsoft-icecaps","post_type":"msr-project","post_date":"2019-08-29 10:40:23","post_modified":"2019-11-05 11:51:07","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/microsoft-icecaps\/","post_excerpt":"With natural language processing rapidly increasing in popularity, more and more tools have become available to the public to build large systems. Some of these tools are intended for general-purpose NLP, while others focus on specific domains such as language modeling and text generation. However, few are designed to target conversational scenarios and the specific needs they entail. Microsoft Icecaps was created to offer researchers and developers an open-source toolkit with a focus on conversational&hellip;","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/604608"}]}},{"ID":393245,"post_title":"Conversational Intelligence","post_name":"conversational-intelligence","post_type":"msr-project","post_date":"2017-07-05 10:01:45","post_modified":"2017-11-15 13:39:25","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/conversational-intelligence\/","post_excerpt":"Intelligent agents that can handle human language play a growing role in personalized, ubiquitous computing and the everyday use of devices. Agents need to be able to communicate and collaborate with humans in ways that are seamless and natural, and to be able to learn new behaviors, concepts, and relationships as first-class operations. In other words, our devices need to be able to converse with us. In this project, Microsoft Research AI teams are interested&hellip;","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/393245"}]}},{"ID":171447,"post_title":"Data-Driven Conversation","post_name":"data-driven-conversation","post_type":"msr-project","post_date":"2015-03-19 17:13:58","post_modified":"2019-08-19 10:40:23","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/data-driven-conversation\/","post_excerpt":"This project aims to enable people to converse with their devices. We are trying to teach devices to engage with humans using human language in ways that appear seamless and natural to humans. Our research focuses on statistical methods by which devices can learn from human-human conversational interactions and can situate responses in the verbal context and in physical or virtual environments. Natural and Engaging Agents that process human language will play a growing role&hellip;","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/171447"}]}},{"ID":171065,"post_title":"Recurrent Neural Networks for Language Processing","post_name":"recurrent-neural-networks-for-language-processing","post_type":"msr-project","post_date":"2012-11-23 11:45:31","post_modified":"2019-08-19 14:55:37","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/recurrent-neural-networks-for-language-processing\/","post_excerpt":"This project focuses on advancing the state-of-the-art in language processing with recurrent neural networks. We are currently applying these to language modeling, machine translation, speech recognition, language understanding and meaning representation. A special interest in is adding side-channels of information as input, to model phenomena which are not easily handled in other frameworks. A toolkit for doing RNN language modeling with side-information is in the associated download. Sample word vectors for use with this toolkit&hellip;","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/171065"}]}}]},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/238042","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":1,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/238042\/revisions"}],"predecessor-version":[{"id":514181,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/238042\/revisions\/514181"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=238042"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=238042"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=238042"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=238042"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=238042"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=238042"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=238042"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=238042"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=238042"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=238042"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=238042"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=238042"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=238042"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}