{"id":398939,"date":"2017-07-11T11:59:05","date_gmt":"2017-07-11T18:59:05","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-research-item&#038;p=398939"},"modified":"2018-10-16T20:03:35","modified_gmt":"2018-10-17T03:03:35","slug":"evaluation-language-model-using-clustered-model-backoff","status":"publish","type":"msr-research-item","link":"https:\/\/www.microsoft.com\/en-us\/research\/publication\/evaluation-language-model-using-clustered-model-backoff\/","title":{"rendered":"Evaluation of a language model using a clustered model backoff"},"content":{"rendered":"<div class=\"container-fluid body-content\">\n<article class=\"author-page\" data-bind=\"page: {\n         id: 'detail',\n         sourceCache: true,\n         sourceOnShow: 'app\/pages\/detail.cshtml',\n         params: { id: '' }\n     }\"><\/p>\n<div class=\"id-wrapper\" data-bind=\"page: { id: '?', beforeShow: detailPage.beforeDetailPageShown, params: ['jump'] }\">\n<div class=\"pure-g container entity-detail profile-page\">\n<article class=\"container profile-page detail\">\n<section class=\"left-right-comp detail-top\">\n<div class=\"profile-card card paper\">\n<div class=\"pure-g\">\n<div class=\"pure-u-1 pure-u-md-7-8 profile-content\">\n<div>\n<section class=\"paper-abstract\" data-bind=\"with: data.abstract\">\n<p data-bind=\"text: $data\">Describes and evaluates a language model using word classes that have been automatically generated from a word clustering algorithm. Class-based language models have been shown to be effective for rapid adaptation, training on small datasets, and reduced memory usage. In terms of model perplexity, prior work has shown diminished returns for class-based language models constructed using very large training sets. This paper describes a method of using a class model as a backoff to a bigram model which produced significant benefits even when trained from a large text corpus. Tests results on the Whisper continuous speech recognition system show that, for a given word error rate, the clustered bigram model uses 2\/3 fewer parameters compared to a standard bigram model using unigram backoff.<\/p>\n<\/section>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/section>\n<\/article>\n<\/div>\n<\/div>\n<\/article>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>Describes and evaluates a language model using word classes that have been automatically generated from a word clustering algorithm. Class-based language models have been shown to be effective for rapid adaptation, training on small datasets, and reduced memory usage. In terms of model perplexity, prior work has shown diminished returns for class-based language models constructed [&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":"Fourth International Conference on Spoken Language, 1996. ICSLP 96. Proceedings.","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":"Fourth International Conference on Spoken Language, 1996. ICSLP 96. Proceedings.","msr_doi":"10.1109\/ICSLP.1996.607136","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"","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":"1996-10-03","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"http:\/\/ieeexplore.ieee.org\/document\/607136\/","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":0,"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":[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-398939","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-human-language-technologies","msr-locale-en_us"],"msr_publishername":"","msr_edition":"Fourth International Conference on Spoken Language, 1996. ICSLP 96. Proceedings.","msr_affiliation":"","msr_published_date":"1996-10-03","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":"","msr_publicationurl":"http:\/\/ieeexplore.ieee.org\/document\/607136\/","msr_doi":"10.1109\/ICSLP.1996.607136","msr_publication_uploader":[{"type":"url","title":"http:\/\/ieeexplore.ieee.org\/document\/607136\/","viewUrl":false,"id":false,"label_id":0},{"type":"doi","title":"10.1109\/ICSLP.1996.607136","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":"http:\/\/ieeexplore.ieee.org\/document\/607136\/"}],"msr-author-ordering":[{"type":"user_nicename","value":"johnmil","user_id":32370,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=johnmil"},{"type":"user_nicename","value":"fil","user_id":31811,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=fil"}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/398939","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\/398939\/revisions"}],"predecessor-version":[{"id":398942,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/398939\/revisions\/398942"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=398939"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=398939"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=398939"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=398939"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=398939"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=398939"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=398939"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=398939"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=398939"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=398939"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=398939"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=398939"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=398939"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}