{"updated":"2025-01-19T07:36:00.959930+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00241623","sets":["1164:5159:11541:11870"]},"path":["11870"],"owner":"44499","recid":"241623","title":["文単位音声要約のためのデータセット構築とEnd-to-Endモデルの検討"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-12-05"},"_buckets":{"deposit":"5ad26b0f-0197-46c3-9605-050b2c433633"},"_deposit":{"id":"241623","pid":{"type":"depid","value":"241623","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"文単位音声要約のためのデータセット構築とEnd-to-Endモデルの検討","author_link":["665534","665530","665532","665533","665536","665531","665535"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"文単位音声要約のためのデータセット構築とEnd-to-Endモデルの検討"},{"subitem_title":"Constructing Datasets for Sentence-wise Speech Summarization and Exploring End-to-End Modeling","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"音声認識","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2024-12-05","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"日本電信電話株式会社"},{"subitem_text_value":"日本電信電話株式会社"}]},"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/241623/files/IPSJ-SLP24154003.pdf","label":"IPSJ-SLP24154003.pdf"},"date":[{"dateType":"Available","dateValue":"2026-12-05"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-SLP24154003.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":"7e0a3aff-19d5-4148-955c-54c06bfd1537","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"松浦, 孝平"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"芦原, 孝典"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"森谷, 崇史"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"三村, 正人"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"叶, 高朋"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"小川, 厚徳"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"デルクロア, マーク"}],"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":"本研究では,発話文書からテキスト要約を文ごとに生成する新たな手法 “Sentence-wise Speech Summarization (Sen-SSum)” を提案する.Sen-SSum は,自動音声認識(ASR)のリアルタイム性と,音声要約の簡潔さを両立させるアプローチである.本手法を検証するために,Sen-SSum 用の 2 つのデータセット “Mega-SSum” および “CSJ-SSum” を構築する.これらのデータセットを用い,2 種類の Transformer ベースのモデルの性能を評価する.1 つ目は ASR と高性能なテキスト要約モデルを組み合わせたカスケードモデル,2 つ目は音声を直接テキスト要約へ変換する End-to-End(E2E)モデルである.E2E モデルは計算効率の観点から魅力的であるが,カスケードモデルと比較して性能が劣るという課題がある.そこで,カスケードモデルにより生成された擬似要約をもちいて E2E モデルを学習することで,カスケードモデルの強力な言語知識を E2E モデルへ蒸留することを提案する.評価実験により,本手法は E2E モデルの要約精度を両データセットにおいて効果的に向上することを示した.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"7","bibliographic_titles":[{"bibliographic_title":"研究報告音声言語情報処理(SLP)"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2024-12-05","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"3","bibliographicVolumeNumber":"2024-SLP-154"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"created":"2025-01-19T01:46:20.225104+00:00","id":241623,"links":{}}