作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2019, Vol. 45 ›› Issue (6): 211-217. doi: 10.19678/j.issn.1000-3428.0051060

• 人工智能及识别技术 • 上一篇    下一篇

基于词对建模的句子对齐研究

丁颖,李军辉,周国栋   

  1. 苏州大学 自然语言处理实验室,江苏 苏州 215006
  • 收稿日期:2018-04-02 出版日期:2019-06-15 发布日期:2019-06-15
  • 作者简介:丁颖(1994—),女,硕士,主研方向为机器翻译;李军辉,副教授、博士;周国栋,教授、博士。
  • 基金资助:

    国家自然科学基金(61401295)。

Research on sentence alignment based on modeling word pairs

DING Ying,LI Junhui,ZHOU Guodong   

  1. Natural Language Processing Lab,Soochow University,Suzhou,Jiangsu 215006,China
  • Received:2018-04-02 Online:2019-06-15 Published:2019-06-15

摘要:

句子对齐是将源文本中的句子映射到目标文本中对应翻译的过程。在神经网络的框架下,基于相互对齐的源端和目标端句子中包含大量相互对齐的单词,提出一种句子对齐方法。使用门关联网络捕获源端句子和目标端句子词对之间的语义关系,并通过语义关系来确定源端句子和目标端句子是否对齐。对非单调文本进行对齐评估,结果表明,该方法F1值达到93.8%,有效提高了句子对齐的准确率。

关键词: 句子对齐, 词对, 双向循环神经网络, 门关联网络, 语义关系

Abstract:

Sentence alignment is a process mapping sentences in the source text to their counterparts in the target text.Within the framework of neural network,this paper proposes a sentence alignment method,on the basis that the aligned source sentence and target sentence pair contains a large number of aligned words.The Gated Relevance Network (GRN) is used to capture the semantic interaction between the source sentence and the target sentence pair,and the semantic interaction is used to determine whether the source sentence and the target sentence are aligned.The alignment evaluation of non-monotonic text shows that the F1 value of the method reaches 93.8%,which effectively improves the accuracy of sentence alignment.

Key words: sentence alignment, word pairs, Bidirectional Recurrent Neural Network(Bi-RNN), Gated Relevance Network(GRN), semantic interaction

中图分类号: