汉维/维汉统计机器翻译中若干问题研究  

Research on aspects of statistical machine translation between Chinese and Uyghur

作　　者：徐春[1] 杨勇[2] 董兴华[3] 

机构地区：[1]新疆财经大学计算机科学与工程学院,乌鲁木齐830012 [2]新疆师范大学计算机科学技术学院,乌鲁木齐830054 [3]中国科学院新疆理化技术研究所,乌鲁木齐830011

出　　处：《计算机工程与应用》2011年第35期150-154,167,共6页Computer Engineering and Applications

基　　金：中国科学院西部行动计划高新技术基金资助项目(No.KGCX2-YW-507)

摘　　要：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。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 results.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.

关 键 词：汉维 维汉 词对齐 一致性 句法结构 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]