{"links":{},"id":56894,"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00056894","sets":["1164:5159:5174:5177"]},"path":["5177"],"owner":"1","recid":"56894","title":["会議音声の自動話題分割による単語辞書と言語モデルの適応"],"pubdate":{"attribute_name":"公開日","attribute_value":"2006-07-08"},"_buckets":{"deposit":"4dc771fc-2cdb-4e2d-b07c-0ce21e599213"},"_deposit":{"id":"56894","pid":{"type":"depid","value":"56894","revision_id":0},"owners":[1],"status":"published","created_by":1},"item_title":"会議音声の自動話題分割による単語辞書と言語モデルの適応","author_link":["0","0"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"会議音声の自動話題分割による単語辞書と言語モデルの適応"},{"subitem_title":"Vocabulary and Language Model Adaptation based on Automatic Topic Segmentation of Meetings","subitem_title_language":"en"}]},"item_type_id":"4","publish_date":"2006-07-08","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"京都大学情報学研究科知能情報学専攻"},{"subitem_text_value":"京都大学情報学研究科知能情報学専攻"},{"subitem_text_value":"京都大学情報学研究科知能情報学専攻"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"School of Infomatics, Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"School of Infomatics, Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"School of Infomatics, Kyoto University","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/56894/files/IPSJ-SLP06062012.pdf"},"date":[{"dateType":"Available","dateValue":"2008-07-08"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP06062012.pdf","filesize":[{"value":"289.3 kB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"22"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"f1ef0f35-26fc-4743-b063-5b8a568a4127","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2006 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"根本, 雄介"},{"creatorName":"秋田, 祐哉"},{"creatorName":"河原, 達也"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Yusuke, Nemoto","creatorNameLang":"en"},{"creatorName":"Yuya, Akita","creatorNameLang":"en"},{"creatorName":"Tatsuya, Kawahara","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10442647","subitem_source_identifier_type":"NCID"}]},"item_4_textarea_12":{"attribute_name":"Notice","attribute_value_mlt":[{"subitem_textarea_value":"SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc."}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_18gh","resourcetype":"technical report"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"広範な話題からなる会議音声を話題単位に自動分割し，得られた話題ごとに単語辞書と言語モデルの適応を行う手法を提案する．音声認識結果に対してPLSA（Probabilistic Latent Semantic Analysis）を適用して，話題を表す特徴ベクトルに変換し，その類似度に基づいて話題分割を行う．そして，話題ごとに類似したテキストを収集して，単語辞書を更新するとともにN-gram 言語モデルの適応を行う．衆議院予算委員会の音声で評価を行った結果，提案手法により単語辞書・言語モデルの適応を行うことで，ベースラインから未知語率を約25%，テストセットパープレキシティを約9%削減することができた．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"We address a vocabulary and language model adaptation method based on topic segmentation of meetings that include various topics. The ASR result is segmented based on the similarity among the feature vectors that were extracted with PLSA (Probabilistic Latent Semantic Analysis). The relevant texts (newspaper articles) for each topic segment are retrieved. The vocabulary and N-gram language model are updated with this retrieved texts. Experimental evaluation on a meeting of the Lower House Budget Committee showed that the proposed model adaptation based on topic segmentation reduced the test-set OOV rate and perplexity.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"68","bibliographic_titles":[{"bibliographic_title":"情報処理学会研究報告音声言語情報処理（SLP）"}],"bibliographicPageStart":"63","bibliographicIssueDates":{"bibliographicIssueDate":"2006-07-08","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"73(2006-SLP-062)","bibliographicVolumeNumber":"2006"}]},"relation_version_is_last":true,"weko_creator_id":"1"},"created":"2025-01-18T23:20:12.676287+00:00","updated":"2025-01-22T04:45:53.160819+00:00"}