{"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00221123","sets":["6504:11035:11043"]},"path":["11043"],"owner":"44499","recid":"221123","title":["特許文書構造を利用したBERTによる事前学習"],"pubdate":{"attribute_name":"公開日","attribute_value":"2022-02-17"},"_buckets":{"deposit":"ffa3aed4-3e52-4a1e-bb25-b0f1664eb9b4"},"_deposit":{"id":"221123","pid":{"type":"depid","value":"221123","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"特許文書構造を利用したBERTによる事前学習","author_link":["578759","578760","578764","578763","578761","578762"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"特許文書構造を利用したBERTによる事前学習"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"人工知能と認知科学","subitem_subject_scheme":"Other"}]},"item_type_id":"22","publish_date":"2022-02-17","item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_22_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"同志社大"},{"subitem_text_value":"同志社大"},{"subitem_text_value":"同志社大"},{"subitem_text_value":"トランスコスモス株式会社"},{"subitem_text_value":"トランスコスモス株式会社"},{"subitem_text_value":"同志社大"}]},"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/221123/files/IPSJ-Z84-1W-01.pdf","label":"IPSJ-Z84-1W-01.pdf"},"date":[{"dateType":"Available","dateValue":"2022-10-22"}],"format":"application/pdf","filename":"IPSJ-Z84-1W-01.pdf","filesize":[{"value":"556.6 kB"}],"mimetype":"application/pdf","accessrole":"open_date","version_id":"3e24458e-fa4e-4931-9097-3de1c6ff6eee","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2022 by the Information Processing Society of Japan"}]},"item_22_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"湯浅, 亮也"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"谷, 和樹"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"田村, 晃裕"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"伊藤, 和真"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"大林, 弘明"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"加藤, 恒夫"}],"nameIdentifiers":[{}]}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_5794","resourcetype":"conference paper"}]},"item_22_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN00349328","subitem_source_identifier_type":"NCID"}]},"item_22_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年，様々な自然言語処理のタスクで，汎用的な分散表現を事前学習するBERTを活用することで最高精度が達成されている．従来のBERTモデルは文単位でトークンのまとまりを捉え，文章における各文の位置づけは考慮しない．一方，特許文書は，段落単位でまとめられて記述されており，【背景技術】や【発明の概要】などの見出しラベルにより構造化されている．そこで本研究では，段落単位で処理することで段落単位のまとまりを捉え，見出しラベルの情報を取り入れて学習を行う，特許文書のためのBERTによる事前学習手法を提案する．特許文書のクラスタリングの実験を行い，提案手法の方が従来のBERTよりも高いクラスタリング精度を実現できることを確認した．","subitem_description_type":"Other"}]},"item_22_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"774","bibliographic_titles":[{"bibliographic_title":"第84回全国大会講演論文集"}],"bibliographicPageStart":"773","bibliographicIssueDates":{"bibliographicIssueDate":"2022-02-17","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"1","bibliographicVolumeNumber":"2022"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":221123,"updated":"2025-01-19T14:17:23.494240+00:00","links":{},"created":"2025-01-19T01:21:08.227787+00:00"}