汉维/维汉统计机器翻译中若干问题研究

摘要

For the characteristics of large morphological difference between Chinese and Uyghur, with the help of open-source moses toolkit,this paper compares the various translation model in it and analyzes the relevant experimental re-sults.Based on this a variety of factors which may affect the performance of translation between Chinese and Uyghur is deeply studied,including related issues with words alignment,consistency of subject-predicate-tenses in Chinese-Uyghur translation, out-of-vocabulary in Uyghur-Chinese translation, different syntactic structure.Finally,some suggestions are given that how to improve the performance of translation between Chinese and Uyghur.%针对汉语和维吾尔语形态差别较大的特点,借助开源的Moses工具箱,通过各种翻译模型的对比及相关实验结果的分析,深入探讨了对汉维/维汉翻译有影响的各种因素,包括词对齐问题,汉维翻译中主语、谓语中心词、时态等的一致性问题,维汉翻译中OOV的问题,汉维句法结构差异问题.最后给出了提高汉维/维汉统计翻译性能的一些建议.