{"id":238006,"updated":"2025-01-19T08:43:35.025320+00:00","links":{},"created":"2025-01-19T01:40:58.331395+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00238006","sets":["581:11492:11501"]},"path":["11501"],"owner":"44499","recid":"238006","title":["Bilingual Corpus Mining and Multistage Fine-tuning for Improving Machine Translation of Lecture Transcripts"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-08-15"},"_buckets":{"deposit":"68b07fb0-ab05-46d1-b479-0f5ccbb2d42d"},"_deposit":{"id":"238006","pid":{"type":"depid","value":"238006","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"Bilingual Corpus Mining and Multistage Fine-tuning for Improving Machine Translation of Lecture Transcripts","author_link":["651696","651703","651704","651701","651702","651698","651695","651697","651699","651700"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Bilingual Corpus Mining and Multistage Fine-tuning for Improving Machine Translation of Lecture Transcripts"},{"subitem_title":"Bilingual Corpus Mining and Multistage Fine-tuning for Improving Machine Translation of Lecture Transcripts","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"[一般論文] lecture translation, parallel corpus mining, machine translation, sentence alignment","subitem_subject_scheme":"Other"}]},"item_type_id":"2","publish_date":"2024-08-15","item_2_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"National Institute of Information and Communications Technology／Kyoto University"},{"subitem_text_value":"National Institute of Information and Communications Technology"},{"subitem_text_value":"Kyoto University"},{"subitem_text_value":"National Institute of Information and Communications Technology"},{"subitem_text_value":"National Institute of Informatics／Kyoto University"}]},"item_2_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"National Institute of Information and Communications Technology / Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"National Institute of Information and Communications Technology","subitem_text_language":"en"},{"subitem_text_value":"Kyoto University","subitem_text_language":"en"},{"subitem_text_value":"National Institute of Information and Communications Technology","subitem_text_language":"en"},{"subitem_text_value":"National Institute of Informatics / Kyoto University","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"eng"}]},"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/238006/files/IPSJ-JNL6508009.pdf","label":"IPSJ-JNL6508009.pdf"},"date":[{"dateType":"Available","dateValue":"2026-08-15"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-JNL6508009.pdf","filesize":[{"value":"1.2 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"0","billingrole":"5"},{"tax":["include_tax"],"price":"0","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"8"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"0bb3f1ff-09ca-43b5-9b40-68633040b64a","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 by the Information Processing Society of Japan"}]},"item_2_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Haiyue, Song"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Raj, Dabre"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Chenhui, Chu"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Atsushi, Fujita"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Sadao, Kurohashi"}],"nameIdentifiers":[{}]}]},"item_2_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Haiyue, Song","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Raj, Dabre","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Chenhui, Chu","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Atsushi, Fujita","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Sadao, Kurohashi","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_2_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AN00116647","subitem_source_identifier_type":"NCID"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_6501","resourcetype":"journal article"}]},"item_2_publisher_15":{"attribute_name":"公開者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"item_2_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"1882-7764","subitem_source_identifier_type":"ISSN"}]},"item_2_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"Lecture transcript translation helps learners understand online courses; however, building a high-quality lecture machine translation system lacks publicly available parallel corpora. To address this, we examine a framework for parallel corpus mining, which provides a quick and effective way to mine a parallel corpus from publicly available lectures on Coursera. To create the parallel corpora, we propose a dynamic programming based sentence alignment algorithm which leverages the cosine similarity of machine-translated sentences. The sentence alignment F1 score reaches 96%, which is higher than using the BERTScore, LASER, or sentBERT methods. For both English-Japanese and English-Chinese lecture translations, we extracted parallel corpora of approximately 50,000 lines and created development and test sets through manual filtering for benchmarking translation performance. Through machine translation experiments, we show that the mined corpora enhance the quality of lecture transcript translation when used in conjunction with out-of-domain parallel corpora via multistage fine-tuning. Furthermore, this study also suggests guidelines for gathering and cleaning corpora, mining parallel sentences, cleaning noise in the mined data, and creating high-quality evaluation splits. For the sake of reproducibility, we have released the corpora as well as the code to create them.\n------------------------------\nThis is a preprint of an article intended for publication Journal of\nInformation Processing(JIP). This preprint should not be cited. This\narticle should be cited as: Journal of Information Processing Vol.32(2024) (online)\nDOI　http://dx.doi.org/10.2197/ipsjjip.32.628\n------------------------------","subitem_description_type":"Other"}]},"item_2_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"Lecture transcript translation helps learners understand online courses; however, building a high-quality lecture machine translation system lacks publicly available parallel corpora. To address this, we examine a framework for parallel corpus mining, which provides a quick and effective way to mine a parallel corpus from publicly available lectures on Coursera. To create the parallel corpora, we propose a dynamic programming based sentence alignment algorithm which leverages the cosine similarity of machine-translated sentences. The sentence alignment F1 score reaches 96%, which is higher than using the BERTScore, LASER, or sentBERT methods. For both English-Japanese and English-Chinese lecture translations, we extracted parallel corpora of approximately 50,000 lines and created development and test sets through manual filtering for benchmarking translation performance. Through machine translation experiments, we show that the mined corpora enhance the quality of lecture transcript translation when used in conjunction with out-of-domain parallel corpora via multistage fine-tuning. Furthermore, this study also suggests guidelines for gathering and cleaning corpora, mining parallel sentences, cleaning noise in the mined data, and creating high-quality evaluation splits. For the sake of reproducibility, we have released the corpora as well as the code to create them.\n------------------------------\nThis is a preprint of an article intended for publication Journal of\nInformation Processing(JIP). This preprint should not be cited. This\narticle should be cited as: Journal of Information Processing Vol.32(2024) (online)\nDOI　http://dx.doi.org/10.2197/ipsjjip.32.628\n------------------------------","subitem_description_type":"Other"}]},"item_2_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographic_titles":[{"bibliographic_title":"情報処理学会論文誌"}],"bibliographicIssueDates":{"bibliographicIssueDate":"2024-08-15","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"8","bibliographicVolumeNumber":"65"}]},"relation_version_is_last":true,"weko_creator_id":"44499"}}