{"id":163799,"date":"2011-03-01T00:00:00","date_gmt":"2011-03-01T00:00:00","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/msr-research-item\/variational-approximation-of-long-span-language-models-for-lvcsr\/"},"modified":"2018-10-16T20:00:49","modified_gmt":"2018-10-17T03:00:49","slug":"variational-approximation-of-long-span-language-models-for-lvcsr","status":"publish","type":"msr-research-item","link":"https:\/\/www.microsoft.com\/en-us\/research\/publication\/variational-approximation-of-long-span-language-models-for-lvcsr\/","title":{"rendered":"Variational Approximation of Long-Span Language Models for LVCSR"},"content":{"rendered":"<div class=\"asset-content\">\n<p>Long-span language models that capture syntax and semantics are seldom used in the first<br \/>\npass of large vocabulary continuous speech recognition systems due to the prohibitive<br \/>\nsearch-space of sentencehypotheses. Instead, an N-best list of hypotheses is created<br \/>\nusing tractable n-gram models, and rescored using the long-span models. It is shown in<br \/>\nthis paper that computationally tractable variational approximations of the long-span<br \/>\nmodels are a better choice than standard n-gram models for first pass decoding. They not<br \/>\nonly result in a better first pass output, but also produce a lattice with a lower oracle word<br \/>\nerror rate, and rescoring the N-best list from such lattices with the long-span models<br \/>\nrequires a smaller N to attain the same accuracy. Empirical results on the WSJ, MIT<br \/>\nLectures, NIST 2007 Meeting Recognition and NIST 2001 Conversational Telephone<br \/>\nRecognition data sets are presented to support these claims.<\/p>\n<\/div>\n<p><!-- .asset-content --><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Long-span language models that capture syntax and semantics are seldom used in the first pass of large vocabulary continuous speech recognition systems due to the prohibitive search-space of sentencehypotheses. Instead, an N-best list of hypotheses is created using tractable n-gram models, and rescored using the long-span models. It is shown in this paper that computationally [&hellip;]<\/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":[],"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-163799","msr-research-item","type-msr-research-item","status-publish","hentry","msr-locale-en_us"],"msr_publishername":"IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)","msr_edition":"","msr_affiliation":"","msr_published_date":"2011-03-01","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":"220564","msr_publicationurl":"http:\/\/www.superlectures.com\/icassp2011\/lecture.php?lang=en&id=226","msr_doi":"","msr_publication_uploader":[{"type":"file","title":"ICASSP-2011.pdf","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2011\/03\/ICASSP-2011.pdf","id":220564,"label_id":0},{"type":"url","title":"http:\/\/www.superlectures.com\/icassp2011\/lecture.php?lang=en&id=226","viewUrl":false,"id":false,"label_id":0}],"msr_related_uploader":"","msr_attachments":[{"id":0,"url":"http:\/\/www.superlectures.com\/icassp2011\/lecture.php?lang=en&#038;id=226"},{"id":220564,"url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2011\/03\/ICASSP-2011.pdf"}],"msr-author-ordering":[{"type":"user_nicename","value":"anoopd","user_id":31047,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=anoopd"}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[],"msr_project":[171065],"publication":[],"video":[],"download":[],"msr_publication_type":"inproceedings","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/163799"}],"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\/163799\/revisions"}],"predecessor-version":[{"id":518446,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/163799\/revisions\/518446"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=163799"}],"wp:term":[{"taxonomy":"msr-content-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-content-type?post=163799"},{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=163799"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=163799"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=163799"},{"taxonomy":"msr-product-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-product-type?post=163799"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=163799"},{"taxonomy":"msr-platform","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-platform?post=163799"},{"taxonomy":"msr-download-source","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-download-source?post=163799"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=163799"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=163799"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=163799"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=163799"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=163799"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=163799"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}