An Empirical Study of Incorporating Pseudo Data into Grammatical Error Correction
長岡技術科学大学
自然言語処理研究室
文献紹介(2019-09-30)
An Empirical Study of Incorporating Pseudo Data into Grammatical Error Correction
https://arxiv.org/pdf/1909.00502.pdf
JOINT Joint Training vs Pre-training 8 PRETRAIN learner data pseudo data learner data pseudo data Pre-training Training Fine-tuning 擬似誤りデータを訓練するための 2つのパターンを検証
References 14 ● BACKTRANS(NOISE) ○ Noising and Denoising Natural Language: Diverse Backtranslation for Grammar Correction (2018) ● BACKTRANS(SAMPLE) ○ Understanding Back-Translation at Scale (2018) ● DIRECT NOISE ○ Improving Grammatical Error Correction via Pre-Training a Copy-Augmented Architecture with Unlabeled Data (2019) ● Synthetic Spelling Error(SSE) ○ Corpora Generation for Grammatical Error Correction (2019) ● Right-to-left Re-ranking(R2L) ○ Neural grammatical error correction systems with unsupervised pre-training on synthetic data (2019) ● Sentence-level Error Detection(SED) ○ The AIP-Tohoku System at the BEA-2019 Shared Task (2019)