北京大学学报自然科学版 ›› 2022, Vol. 58 ›› Issue (1): 61-68.DOI: 10.13209/j.0479-8023.2021.111

上一篇    下一篇

融合小句对齐知识的汉英神经机器翻译

苗国义1, 刘明童2, 陈钰枫1, 徐金安1,†, 张玉洁1, 冯文贺3   

  1. 1. 北京交通大学计算机与信息技术学院, 北京 100044 2. 创新工场人工智能工程院, 北京 100080 3. 广东外语外贸大学语言工程与计算实验室, 广州 510420
  • 收稿日期:2021-06-09 修回日期:2021-08-13 出版日期:2022-01-20 发布日期:2022-01-20
  • 通讯作者: 徐金安, E-mail: jaxu(at)bjtu.edu.cn
  • 基金资助:
    国家重点研发计划(2020AAA0108001)、国家自然科学基金(61976015, 61976016, 61876198, 61370130)和广东省基础与应用基础研究基金(2020A1515011056)资助

Incorporating Clause Alignment Knowledge into Chinese-English Neural Machine Translation

MIAO Guoyi1, LIU Mingtong2, CHEN Yufeng1, XU Jin’an1,†, ZHANG Yujie1, FENG Wenhe3   

  1. 1. School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044 2. Sinovation Ventures AI Institute, Beijing, 100080 3. Laboratory of Language Engineering and Computing, Guangdong University of Foreign Studies, Guangzhou 510420
  • Received:2021-06-09 Revised:2021-08-13 Online:2022-01-20 Published:2022-01-20
  • Contact: XU Jin'an, E-mail: jaxu(at)bjtu.edu.cn

摘要:

针对当前神经机器翻译在捕捉复杂句内小句间的语义和结构关系方面存在不足, 导致复杂句长文本翻译的篇章连贯性不佳的问题, 提出一种融合小句对齐知识的汉英神经机器翻译方法。首先提出手工和自动相结合的标注方案, 构建大规模小句对齐的汉英平行语料库, 为模型训练提供丰富的小句级别的汉英双语对齐知识; 然后设计一种基于小句对齐学习的神经机器翻译模型, 通过融合小句对齐知识, 增强模型学习复杂句内小句间语义结构关系的能力。在WMT17, WMT18和WMT19汉英翻译任务中的实验表明, 所提出的方法可以有效地提升神经机器翻译的性能。进一步的评测分析显示, 所提方法能有效地提高汉英神经机器翻译在复杂句翻译上的篇章连贯性。

关键词: 神经机器翻译, 小句对齐, 结构关系, 篇章连贯性

Abstract:

Currently, neural machine translation (NMT) is insufficient in capturing the semantic and structural relationships between clauses in complex sentences, which often results in poor discourse coherence of long and complex sentence translation. To address this problem, the paper proposes a Chinese-English NMT approach by integrating the clause alignment knowledge into NMT. Firstly, a labeling scheme combining manual and automatic annotation is introduced to annotate a large-scale clause aligned Chinese-English parallel corpus that provides rich clause-level Chinese-English bilingual alignment knowledge for model training. Then, a NMT model is designed based on clause alignment learning for enhancing the ability of the model to learn the semantic structure relationships between clauses within complex sentences. Experimental results on WMT17, WMT18 and WMT19 Chinese-English translation tasks demonstrate that proposed method can significantly improve the NMT performance. Evaluation and analysis show that proposed method can effectively improve the discourse coherence of complex sentence in Chinese-English machine translation.

Key words: neural machine translation, clause alignment, structural relationship, discourse coherence