Discriminative Reordering with Chinese Grammatical Relations Features
Discriminative Reordering with Chinese Grammatical Relations Features
复制标题
DOI：
10.3115/1626344.1626351
复制
发表时间：
2009-06
期刊：
影响因子：
--
通讯作者：
Pi-Chuan Chang;Huihsin Tseng;Dan Jurafsky;Christopher D. Manning
中科院分区：
文献类型：
--
作者：
Pi-Chuan Chang;Huihsin Tseng;Dan Jurafsky;Christopher D. Manning
语法结构中文的普遍性以不同的单词顺序转化为英语是翻译困难的重要原因。尽管以前的工作使用短语结构解析来处理此类有序问题，但我们介绍了一组更丰富的中国语法关系，描述了单词之间更加语义上抽象的关系。使用这些中国语法关系，我们改善了短语取向分类器（由Zens和Ney（2006）介绍），该分类器通过添加在中国打字依赖项上设计的路径特征来决定两个短语的顺序。然后，我们将短语方向分类器的对数概率应用于基于短语的MT系统中的额外功能，并在三个测试集上获得显着的BLEU点提升：MT02（+0.59），MT03（+1.00）（+1.00）和MT05（+0.77777777777777777777777777777777777777777777777777777777777777777777777777777） ）。我们的中国语法关系也可能对其他NLP任务有用。
The prevalence in Chinese of grammatical structures that translate into English in different word orders is an important cause of translation difficulty. While previous work has used phrase-structure parses to deal with such ordering problems, we introduce a richer set of Chinese grammatical relations that describes more semantically abstract relations between words. Using these Chinese grammatical relations, we improve a phrase orientation classifier (introduced by Zens and Ney (2006)) that decides the ordering of two phrases when translated into English by adding path features designed over the Chinese typed dependencies. We then apply the log probability of the phrase orientation classifier as an extra feature in a phrase-based MT system, and get significant BLEU point gains on three test sets: MT02 (+0.59), MT03 (+1.00) and MT05 (+0.77). Our Chinese grammatical relations are also likely to be useful for other NLP tasks.