计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (36): 142-144.DOI: 10.3778/j.issn.1002-8331.2009.36.042

• 数据库、信号与信息处理 • 上一篇    下一篇

句法调序的统计机器翻译方法研究

孙广范,宋金平,肖 健,袁 琦   

  1. 中国电子信息产业发展研究院,北京 100044
  • 收稿日期:2008-07-11 修回日期:2008-11-17 出版日期:2009-12-21 发布日期:2009-12-21
  • 通讯作者: 孙广范

Research on approaches of syntactic reordering for statistical machine translation

SUN Guang-fan,SONG Jin-ping,XIAO Jian,YUAN Qi   

  1. China Center for Information Industry Development,Beijing 100044,China
  • Received:2008-07-11 Revised:2008-11-17 Online:2009-12-21 Published:2009-12-21
  • Contact: SUN Guang-fan

摘要: 为解决基于短语统计机器翻译存在的调序能力不足的问题,尝试利用句法分析器对基于短语统计机器翻译的输入汉语句子进行句法分析,然后利用转换器进行调序操作,并对部分类型短语进行预先翻译,然后再利用基于短语统计机器翻译的解码器进行翻译。重点测试了汉语中“的”字引导的复杂定语调序、介词短语、特定搭配短语、方位词短语的调序及预翻译产生的效果。实验结果表明,这些调序及预翻译操作可以显著地提高基于短语的统计机器翻译的英文译文结果的BLEU值。

关键词: 统计机器翻译, 句法调序, 预先翻译

Abstract: To resolve the problem of the lack of capacity for the reordering in Phrase-Based Statistical Machine Translation(PBSMT),this paper attempts to parse the input Chinese sentences to PBSMT,conduct the reordering of the sentences by a converter,pre-translate some types of phrases,and then translate the output from the above operation by a decoder.This paper focuses on the effects of the reordering for complex attributes linked by“de”,and the effects of the reordering and pre-translation of prepositional phrases,specific collocation phrases and phrases of noun of locality.Experimental results show that the reordering and pre-translation can significantly improve the BLEU values of the English translation by PBSMT.

Key words: statistical machine translation, syntactic reordering, pre-translation

中图分类号: