{"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00211526","sets":["1164:5064:10547:10607"]},"path":["10607"],"owner":"44499","recid":"211526","title":["End-to-end音声認識モデルにおける暗黙的言語情報の置換法"],"pubdate":{"attribute_name":"公開日","attribute_value":"2021-06-11"},"_buckets":{"deposit":"9e399b72-cce9-428c-8796-70e17462a1de"},"_deposit":{"id":"211526","pid":{"type":"depid","value":"211526","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"End-to-end音声認識モデルにおける暗黙的言語情報の置換法","author_link":["537512","537516","537515","537519","537513","537518","537520","537511","537517","537514"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"End-to-end音声認識モデルにおける暗黙的言語情報の置換法"},{"subitem_title":"Language Model replacement method for end-to-end speech recognition which excludes implicit linguistic information","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"一般発表","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2021-06-11","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"豊橋技術科学大学"},{"subitem_text_value":"阿南工業高等専門学校"},{"subitem_text_value":"徳島大学"},{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"豊橋技術科学大学"}]},"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/211526/files/IPSJ-MUS21131017.pdf","label":"IPSJ-MUS21131017.pdf"},"date":[{"dateType":"Available","dateValue":"2023-06-11"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-MUS21131017.pdf","filesize":[{"value":"835.3 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":"21"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"64a9ebad-ca0e-493b-8bb8-e9313c6ee018","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2021 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"森, 大輝"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"太田, 健吾"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"西村, 良太"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"小川, 厚徳"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"北岡, 教英"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Daiki, Mori","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Kengo, Ohta","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Ryota, Nishimura","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Atsunori, Ogawa","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Norihide, Kitaoka","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN10438388","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_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2188-8752","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年，End-to-end 音声認識が従来の DNN-HMM 音声認識と比べ，高速かつ簡潔であることから注目されている．さらに大量のテキストデータによって学習された言語モデルを併用することで，認識精度が向上すると報告されている．本稿では，音声認識モデルと言語モデルの一般的な統合方法とされる Shallow Fusion を応用した新しい言語モデルの統合方法である Language Model Replacement を提案する．提案法では，事前学習済み音声認識モデルと事前学習済み言語モデルを用いる．提案法ではベイズ則に基づき，音声認識モデルに暗黙的に含まれる言語情報を差し替えることが可能となっている．我々の実験では，学術講演音声データを使用して学習された音声認識モデル内部の言語情報を，模擬講演テキストデータで学習した言語モデルによって差し替えた．模擬講演ドメインにおける提案法の CER は Shallow Fusion での認識精度と比較して，1.3 ポイント上回った．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"Recently, end-to-end speech recognition has attracted much attention because it is faster and more concise than conventional DNN-HMM speech recognition. It has also been reported that recognition performance is improved by employing a language model trained with a large amount of text data. Based on these observations, we propose a new language model integration method which we call Language Model Replacement. In our proposed method, we use a pre-trained speech recognition model and a pre-trained language model. In contrast to the Shallow Fusion method, our proposed method can replace the linguistic information implied in the ASR model with independently trained model based on Bayes' rule. In our experiments, the ASR linguistic information implicitly trained using the Japanese language Academic Presentation Speech corpus is replaced with the language model trained using the Japanese language Simulated Public Speech corpus. We then compare ASR performance for Japanese speech recognition tasks using the Character Error Rate (CER). Our proposed Language Model Replacement method achieved 1.3 percent lower CER in comparison to the Shallow Fusion method.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"6","bibliographic_titles":[{"bibliographic_title":"研究報告音楽情報科学（MUS）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2021-06-11","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"17","bibliographicVolumeNumber":"2021-MUS-131"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":211526,"updated":"2025-01-19T17:46:12.551189+00:00","links":{},"created":"2025-01-19T01:12:40.409451+00:00"}