{"links":{},"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00069914","sets":["1164:5159:6009:6133"]},"path":["6133"],"owner":"10","recid":"69914","title":["ベイズ推論を用いた連続音声からの言語モデル学習"],"pubdate":{"attribute_name":"公開日","attribute_value":"2010-07-15"},"_buckets":{"deposit":"f0f99282-02ed-47d1-ba89-84343a7242de"},"_deposit":{"id":"69914","pid":{"type":"depid","value":"69914","revision_id":0},"owners":[10],"status":"published","created_by":10},"item_title":"ベイズ推論を用いた連続音声からの言語モデル学習","author_link":["0","0"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"ベイズ推論を用いた連続音声からの言語モデル学習"},{"subitem_title":"Learning a Language Model from Continuous Speech using Bayesian Inference","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"言語モデル","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2010-07-15","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"},{"subitem_text_value":"京都大学情報学研究科"},{"subitem_text_value":"京都大学情報学研究科"},{"subitem_text_value":"京都大学情報学研究科"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Kyoto University, Graduate School of Informatics","subitem_text_language":"en"},{"subitem_text_value":"Kyoto University, Graduate School of Informatics","subitem_text_language":"en"},{"subitem_text_value":"Kyoto University, Graduate School of Informatics","subitem_text_language":"en"},{"subitem_text_value":"Kyoto University, Graduate School of Informatics","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/69914/files/IPSJ-SLP10082016.pdf"},"date":[{"dateType":"Available","dateValue":"2012-07-15"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP10082016.pdf","filesize":[{"value":"514.8 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":"3677c600-8213-4efd-9da7-321499b66e3f","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2010 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Graham, NEUBIG"},{"creatorName":"三村, 正人"},{"creatorName":"森, 信介"},{"creatorName":"河原, 達也"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Graham, Neubig","creatorNameLang":"en"},{"creatorName":"Masato, Mimura","creatorNameLang":"en"},{"creatorName":"Shinsuke, Mori","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":"本稿ではテキストを用いず，連続音声のみから言語モデルを学習する方法を提案する．音響モデルのみを用いて作成された音素ラティスに対して推論を行い，単語境界と言語モデルを同時に学習する．具体的には，ノンパラメトリックベイズ法に基づく階層的 Pitman-Yor 言語モデルを利用し，パラメータは WFST に基づいたギブスサンプリングで推定する．会議音声を用いた実験において，提案手法によって学習された言語モデルはパープレキシティ及び音声認識の音素誤り率を有意に改善することができた．さらに，ラティス処理と単語単位の獲得が誤り率の改善に貢献していることがわかった．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"This paper proposes a technique for learning a language model directly from continuous speech, without the use of text. Inference is performed over phoneme lattices generated using only acoustic model scores, and word boundaries and a language model are learned simultaneously. A Bayesian non-parametric Hierarchical Pitman-Yor language model is used, and parameters are estimated with WFST-based Gibbs sampling. An experiment was performed using meeting speech, and language models built using the proposed techniques were able to significantly lower ASR phoneme error rates. In addition, lattice processing and word boundary discovery were shown to contribute significantly to this improvement.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告音声言語情報処理（SLP）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2010-07-15","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"16","bibliographicVolumeNumber":"2010-SLP-82"}]},"relation_version_is_last":true,"weko_creator_id":"10"},"updated":"2025-01-21T23:43:10.823790+00:00","created":"2025-01-18T23:29:15.602178+00:00","id":69914}