arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.10707cs.CLcs.LG
收藏

基于预训练跨语言语言模型的多语言语法纠错统一策略

A Unified Strategy for Multilingual Grammatical Error Correction with Pre-trained Cross-Lingual Language Model

Peking University(北京大学);Microsoft Research Asia(微软亚洲研究院);The Chinese University of Hong Kong(香港中文大学)

Xin Sun, Tao Ge, Shuming Ma, Jingjing Li, Furu Wei, Houfeng Wang

摘要:

非英语语法纠错(GEC)的合成数据构建严重依赖于人为设计的特定语言规则,产生有限的纠错模式。在本文中,我们提出了一个通用的和语言无关的多语言GEC策略,它可以训练一个GEC系统有效地为一个新的非英语语言,只有两个易于访问的资源:1)一个预训练的跨语言语言模型(PXLM)和2)英语和语言之间的并行翻译数据。我们的方法创建了不同的并行GEC数据,没有任何语言特定的操作,通过将PXLM生成的非自回归翻译和黄金翻译作为纠错句子对。然后,我们重用PXLM初始化GEC模型,并使用其自身生成的合成数据对其进行预训练,这产生了进一步的改进。我们评估我们的方法在不同语言的GEC的三个公共基准。它在NLPCC 2018任务2数据集(中文)上获得了最先进的结果,并在Falko-Merlin(德国)和RULEC-GEC(俄罗斯)上获得了具有竞争力的性能。进一步的分析表明,我们的数据构建方法是基于规则的方法的补充。

英文摘要:

Synthetic data construction of Grammatical Error Correction (GEC) for non-English languages relies heavily on human-designed and language-specific rules, which produce limited error-corrected patterns. In this paper, we propose a generic and language-independent strategy for multilingual GEC, which can train a GEC system effectively for a new non-English language with only two easy-to-access resources: 1) a pretrained cross-lingual language model (PXLM) and 2) parallel translation data between English and the language. Our approach creates diverse parallel GEC data without any language-specific operations by taking the non-autoregressive translation generated by PXLM and the gold translation as error-corrected sentence pairs. Then, we reuse PXLM to initialize the GEC model and pretrain it with the synthetic data generated by itself, which yields further improvement. We evaluate our approach on three public benchmarks of GEC in different languages. It achieves the state-of-the-art results on the NLPCC 2018 Task 2 dataset (Chinese) and obtains competitive performance on Falko-Merlin (German) and RULEC-GEC (Russian). Further analysis demonstrates that our data construction method is complementary to rule-based approaches.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑