J., Liu, Z. (2020). A Comprehensive Survey of Grammar Error Correc- tion. arXiv preprint arXiv:2005.06600. [Mizumoro 2012] Mizumoto, T., Komachi, M., Nagata, M., & Matsumoto, Y. (2011, November). Mining revision log of language learning SNS for automated Japanese error correction of second language learners. In Proceedings of 5th International Joint Conference on Natural Language Processing (pp. 147-155). [Kiyono 2019] Kiyono, S., Suzuki, J., Mita, M., Mizumoto, T., & Inui, K. (2019). An empirical study of incorporating pseudo data into grammatical error correction. arXiv preprint arXiv:1909.00502. [Koyama 2020] Koyama, A., Kiyuna, T., Kobayashi, K., Arai, M., Komachi, M. (2020, May). Construction of an Eval- uation Corpus for Grammatical Error Correction for Learners of Japanese as a Second Language. In Proceedings of The 12th Language Resources and Evaluation Conference (pp. 204-211). [Grundkiewicz] R. Grundkiewicz et al. “Neural Grammatical Error Correction Sys- tems with Unsupervised Pre- training on Synthetic Data”. In: BEA. 2019, pp. 252–263. [Zhao 2019] W. Zhao et al. “Improving Grammatical Error Correction via Pre- Training a Copy-Augmented Architecture with Unlabeled Data”. In: NAACL. 2019. [Vaswani 2017] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. In Advances in neural information processing systems (pp. 5998-6008). [Devlin 2018] Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2018). Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805. [Maekawa 2014] KikuoMaekawa,MakotoYamazaki,ToshinobuOgiso,TakehikoMaruyama,HidekiOgura,Wakako Kashino, Hanae Koiso, Masaya Yamaguchi, Makiro Tanaka, and Yasuharu Den. Balanced corpus of contemporary written japanese. Language resources and evaluation, Vol. 48, No. 2, pp. 345–371, 2014. [tohoku BERT] https://github.com/cl-tohoku/bert-japanese [ogawa 2019] https://github.com/youichiro/transformer-copy
hideyoshikato
teacherpeterpan
yuta0306
kingqwert
kentaroy47
noppoman
matsui_528
japanstrokeassociation
kurita
usaito
naok615
nehi0615
hirachan
thekraken
hannesfritz
paulrobertlloyd
stephaniewalter
marktimemedia
lara
sstephenson
yeseniaperezcruz
holman
mraible
bryan
3n
![17 /36 要素1 ⽣成元データ ⽣成元データとして複数のドメインを持つデータで実験を⾏い、⽐較検討を⾏う 本研究では、BCCWJ[Maekawa 2014] から抽出した ・BCCWJ-LB・・・図書館書籍ドメイン ・BCCWJ-PB・・・出版書籍](https://files.speakerdeck.com/presentations/748e8aa924a14b3088549b70f759876f/slide_16.jpg){kind=link}
![18 /36 要素2 データ⽣成量 BERT-DAは、BERT[tohoku BERT] の Masked LM によってマスクされたトークンに対する予測トークンを⽤いてい](https://files.speakerdeck.com/presentations/748e8aa924a14b3088549b70f759876f/slide_17.jpg){kind=link}
![21 /36 実験設定︓評価指標 GLEU[Napoles 2015] ・機械翻訳の評価尺度BLEUをGECのために改良した評価指標 ・誤り⽂(src), 正解⽂(tgt), GECモデルによる出⼒⽂(out)の3つの データを使って評価する](https://files.speakerdeck.com/presentations/748e8aa924a14b3088549b70f759876f/slide_20.jpg){kind=link}
![34 /36 参考⽂献 [Wang 2020] Wang, Y., Wang, Y., Liu,](https://files.speakerdeck.com/presentations/748e8aa924a14b3088549b70f759876f/slide_33.jpg){kind=link}
![35 /36 参考⽂献 [Dahlmeier 2012] Daniel Dahlmeier and Hwee Tou](https://files.speakerdeck.com/presentations/748e8aa924a14b3088549b70f759876f/slide_34.jpg){kind=link}
![36 /36 誤りタイプ⼀覧 ▪削除(助詞) (例)「英語がわかる」→「英語[]わかる」 ▪挿⼊(助詞) (例)「英語がわかる」→「英語が[を]わかる」 ▪[置換(助詞)] (例)「英語がわかる」→「英語[に]わかる」 ▪[語彙選択]](https://files.speakerdeck.com/presentations/748e8aa924a14b3088549b70f759876f/slide_35.jpg){kind=link}
