汉英机器翻译中“的”的二义性消除

Removal of the Ambiguity of Chinese “的 De” in English Chinese Machine Translation System

汉语连接结构中的"的(DE)"无处不在,并且可以以多种不同的方式翻译成英语。即使使用语法敏感的翻译模型,其仍然是机器翻译产生偏差的一个主要原因。如何根据语篇语境中的"的(DE)"获得句法、语义的更多信息,从而催生出适当的英语翻译策略?实验描述了汉语NPs中"的(DE)"的分类,构造带注释的实例语料库,然后将其训练成包含语言启发功能的分类器。通过显式标签"的(DE)"结构,使用DE分类器对机器翻译数据进行预处理及重排短语,结果表明此方法为措辞系统提供了显著的MT02(使用BLEU及其翻译软件系统)收益(+1.31),MT03(+0.92)和MT05(+1.57),当翻译模型的应用层次经过如此重新排序后其准确性得到持续提升。

Chinese ＂De＂ could be used widely in linking structures and be translated into different expressions. This causes the key problem in machine translation systems whatever sensitive translation software used. The essay discusses how to get more information of semantics and grammar. We analyze a finer grained classification of Chinese NPs and set up an annotated corpus database and use this classification to process and reconstruct the words. The result shows that this method provides positive benefits by using BLEU for 1.31, 0.92 and 1.57 up respectively. The accuracy and fluency has been greatly enhanced by reconstructing the sequence of Chinese ＂De＂ .