{"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00192707","sets":["1164:5159:9402:9617"]},"path":["9617"],"owner":"44499","recid":"192707","title":["音声中の検索語検出におけるドキュメントの最尤系列化と上位候補の再照合方式による検索時間・精度の改善"],"pubdate":{"attribute_name":"公開日","attribute_value":"2018-12-03"},"_buckets":{"deposit":"dd11854d-1cd7-45b6-a97d-7e2a4c12cb6a"},"_deposit":{"id":"192707","pid":{"type":"depid","value":"192707","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"音声中の検索語検出におけるドキュメントの最尤系列化と上位候補の再照合方式による検索時間・精度の改善","author_link":["450454","450451","450455","450453","450450","450457","450452","450456"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"音声中の検索語検出におけるドキュメントの最尤系列化と上位候補の再照合方式による検索時間・精度の改善"},{"subitem_title":"Improvement of Search Accuracy and Search Time in Query-by-Example by Using Maximum Likelihood Sequence and Rescoring Top Candidates","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"学生ポスターセッション","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2018-12-03","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"岩手県立大学"},{"subitem_text_value":"岩手県立大学"},{"subitem_text_value":"独立行政法人産業技術総合研究所"},{"subitem_text_value":"岩手県立大学"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Iwate Prefecctural University","subitem_text_language":"en"},{"subitem_text_value":"Iwate Prefecctural University","subitem_text_language":"en"},{"subitem_text_value":"National Institute of Advanced Industrial Science and Technology","subitem_text_language":"en"},{"subitem_text_value":"Iwate Prefecctural 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/192707/files/IPSJ-SLP18125010.pdf","label":"IPSJ-SLP18125010.pdf"},"date":[{"dateType":"Available","dateValue":"2020-12-03"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP18125010.pdf","filesize":[{"value":"1.8 MB"}],"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":"f0592212-c311-4cd5-aadb-060df5c248dc","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2018 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"金子, 大祐"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"小嶋, 和徳"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"李, 時旭"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"伊藤, 慶明"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Daisuke, Kaneko","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Kazunori, Kojima","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Shi-Woo, Lee","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Yoshiaki, Itoh","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_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2188-8663","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年,長時間の音声ドキュメント中から特定のシーンを検索する音声中の検索語検出 (STD : Spoken Term Detection) の研究,およびクエリを音声で入力する SQ-STD (Spoken Query STD) の研究が盛んに行われている.DNN (Deep Neural Network) を音声認識に用いることで認識率が大幅に向上することが報告されており,SQ-STD においても DNN を使用することで高精度な検索が可能となった.本稿では SQ-STD における検索時間削減と検索精度向上のための方式を提案する.本稿で提案するドキュメント最優化系列方式では,DNN から得られる事後確率から入力音声の各フレームにおける確率最大値に対応した状態番号を抜き出した系列,最尤系列に変換し予め保持することで,検索の際の計算量の削減を行う.また,検索結果の上位候補の再度照合方式により,検索時間の増加を抑えつつ検索精度の向上を図る.これらの方式のより SQ-STD の検索時間削減および検索精度向上を図る.ドキュメント最尤系列化方式による実験の結果,Posteriorgram 照合と比較して検索語 1 つあたりの検索時間が 90% 以上削減できたとともに,112 GB のメモリ空間使用量が削減できた.また,上位候補の再度 Posteriorgram 照合方式により P@N の評価指標において,検索精度の向上が確認できた.以上の結果から,SQ-STD に対してドキュメント最尤系列化が有効であることを確認した.","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"This paper proposes a maximum likelihood sequence method of retrieval documents obtained from DNN (Deep Neural Network) and a method of rescoring the top candidates of retrieval results again. Improvement search time and accuracy by using the above two methods.","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":"2018-12-03","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"10","bibliographicVolumeNumber":"2018-SLP-125"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":192707,"updated":"2025-01-20T00:02:14.671417+00:00","links":{},"created":"2025-01-19T00:58:24.073005+00:00"}