{"id":168615,"date":"2015-07-01T00:00:00","date_gmt":"2015-07-01T00:00:00","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/msr-research-item\/deltableu-a-discriminative-metric-for-generation-tasks-with-intrinsically-diverse-targets\/"},"modified":"2018-10-16T20:32:35","modified_gmt":"2018-10-17T03:32:35","slug":"deltableu-a-discriminative-metric-for-generation-tasks-with-intrinsically-diverse-targets","status":"publish","type":"msr-research-item","link":"https:\/\/www.microsoft.com\/en-us\/research\/publication\/deltableu-a-discriminative-metric-for-generation-tasks-with-intrinsically-diverse-targets\/","title":{"rendered":"\u02c6\u2020BLEU: A Discriminative Metric for Generation Tasks with Intrinsically Diverse Targets"},"content":{"rendered":"

We introduce Discriminative BLEU (\u2206BLEU), a novel metric for intrinsic evaluation of generated text in tasks that admit a diverse range of possible outputs. Reference strings are scored for quality by human raters on a scale of [\u22121, +1] to weight multi-reference BLEU. In tasks involving generation of conversational responses, \u2206BLEU correlates reasonably with human judgments and outperforms sentence-level and IBM BLEU in terms of both Spearman\u2019s \u03c1 and Kendall\u2019s \u03c4.<\/p>\n","protected":false},"excerpt":{"rendered":"

We introduce Discriminative BLEU (\u2206BLEU), a novel metric for intrinsic evaluation of generated text in tasks that admit a diverse range of possible outputs. Reference strings are scored for quality by human raters on a scale of [\u22121, +1] to weight multi-reference BLEU. In tasks involving generation of conversational responses, \u2206BLEU correlates reasonably with human […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"msr-content-type":[3],"msr-research-highlight":[],"research-area":[13556,13545],"msr-publication-type":[193716],"msr-product-type":[],"msr-focus-area":[],"msr-platform":[],"msr-download-source":[],"msr-locale":[268875],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-168615","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 ACL","msr_affiliation":"","msr_published_date":"2015-07-01","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"445\u2013450","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":"355724","msr_publicationurl":"https:\/\/arxiv.org\/abs\/1506.06863","msr_doi":"","msr_publication_uploader":[{"type":"file","title":"P15-2073","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2015\/07\/P15-2073.pdf","id":355724,"label_id":0},{"type":"url","title":"https:\/\/arxiv.org\/abs\/1506.06863","viewUrl":false,"id":false,"label_id":0}],"msr_related_uploader":"","msr_attachments":[{"id":0,"url":"https:\/\/arxiv.org\/abs\/1506.06863"}],"msr-author-ordering":[{"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":"text","value":"Alessandro Sordoni","user_id":0,"rest_url":false},{"type":"text","value":"Yangfeng Ji","user_id":0,"rest_url":false},{"type":"user_nicename","value":"michauli","user_id":32906,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=michauli"},{"type":"user_nicename","value":"chrisq","user_id":31430,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=chrisq"},{"type":"user_nicename","value":"memitc","user_id":32882,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=memitc"},{"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":[],"msr_project":[393245,171447],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/168615"}],"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":2,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/168615\/revisions"}],"predecessor-version":[{"id":528493,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/168615\/revisions\/528493"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=168615"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=168615"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=168615"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=168615"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=168615"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=168615"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=168615"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=168615"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=168615"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=168615"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=168615"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=168615"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=168615"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=168615"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=168615"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}