{"links":{},"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00057597","sets":["1164:5159:5216:5219"]},"path":["5219"],"owner":"1","recid":"57597","title":["大語彙連続音声認識における認識誤り原因の自動同定"],"pubdate":{"attribute_name":"公開日","attribute_value":"1999-07-23"},"_buckets":{"deposit":"4bb5cf5c-9861-454b-822d-eb2ca4080613"},"_deposit":{"id":"57597","pid":{"type":"depid","value":"57597","revision_id":0},"owners":[1],"status":"published","created_by":1},"item_title":"大語彙連続音声認識における認識誤り原因の自動同定","author_link":["0","0"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"大語彙連続音声認識における認識誤り原因の自動同定"},{"subitem_title":"Automatic Diagnosis of Recognition Errors in Large Vocabulary Continuous Speech Recognition System","subitem_title_language":"en"}]},"item_type_id":"4","publish_date":"1999-07-23","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"京都大学大学院情報学研究科知能情報学専攻"},{"subitem_text_value":"京都大学大学院情報学研究科知能情報学専攻"},{"subitem_text_value":"京都大学大学院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":"Graduate School of Informatics Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Informatics Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"Graduate School of Informatics 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/57597/files/IPSJ-SLP99027006.pdf"},"date":[{"dateType":"Available","dateValue":"2001-07-23"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP99027006.pdf","filesize":[{"value":"650.9 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":"bda8bed1-f45d-481a-b627-ab7d7431a32b","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 1999 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"南條浩輝"},{"creatorName":"李晃伸"},{"creatorName":"河原, 達也"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Hiroaki, Nanjo","creatorNameLang":"en"},{"creatorName":"Akinobu, Lee","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":"音声認識誤りの原因が同定されればシステム改善のための指針を得ることができ、今後の研究の指針やデータ収集時のヒントが得られる。しかし、大語彙連続音声認識においては構成要素となる音響モデル、言語モデルが大規模、かつ統計的モデルであるため認識誤りの原因が何に起因するかを人手で同定するのは容易ではない。本稿では、認識誤りの原因を自動的に同定する手法を提案する。具体的には、正解文を与え、音響モデル、言語モデルから計算されるスコアを認識結果の音響スコア、言語スコアと比較し、認識誤りの原因を音響モデル、言語モデル、探索アルゴリズムのいずれかに同定する。また、一文全体でのスコア比較はモデル改善の指針としては不十分であるため、誤りを含む数個の区間に分割し、区間ごとに原因の同定を行う。探索誤りの場合はそれと提示するにとどめるが、音響モデルが原因であるときは、クラスタリングされて学習されたtriphone、スコアの低いtriphoneを原因と同定し、言語モデルが原因であるときは、低次のN-gramから推定された3-gram，2-gramを原因として同定する。","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"High-quality recognition is required for speech recognition system. If the causes of recognition errors are specified, it is usefull for improvement and next researches. In this paper, we propose the method diagnose errors in the framework of LVCSR. Apply the recognized and correct sentences to an acoustic and a linguistic models, calclate and compare scores, and specify the cause. To be more usefull, separate a sentence to some parts including wrong word, and specify triphones trained together with other triphones and 3-grams and 2-grams not sufficiently trained.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"48","bibliographic_titles":[{"bibliographic_title":"情報処理学会研究報告音声言語情報処理（SLP）"}],"bibliographicPageStart":"41","bibliographicIssueDates":{"bibliographicIssueDate":"1999-07-23","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"64(1999-SLP-027)","bibliographicVolumeNumber":"1999"}]},"relation_version_is_last":true,"weko_creator_id":"1"},"updated":"2025-01-22T04:25:29.180495+00:00","created":"2025-01-18T23:20:45.951878+00:00","id":57597}