基于粘着性模糊规则的维汉机器翻译最大熵调序研究  

Research on max entropy reordering in Uyghur-Chinese machine translation based on tackiness fuzzy rules

作　　者：陈科海[1,2] 周喜[1] 杨雅婷[1] 米成刚[1,2] 

机构地区：[1]中国科学院新疆理化技术研究所,乌鲁木齐830011 [2]中国科学院大学,北京100049

出　　处：《计算机应用研究》2013年第9期2587-2590,2605,共5页Application Research of Computers

基　　金：中国科学院战略性先导科技专项资助项目(XDA06030400);中国科学院"西部之光"人才培养计划西部博士资助项目(XBBS201216);中国科学院西部行动计划项目(KGZD-EW-501)

摘　　要：针对维汉机器翻译中未登录词和译文乱序问题严重的现象,结合维吾尔语粘着性语言特点以及最大熵分类算法,提出了一种基于粘着性模糊规则的维汉机器翻译最大熵调序模型。该模型以最大熵模型为基础,在维吾尔语词级别构建粘着性规则约束,从训练语料中提取更加有效的调序规则来指导翻译解码过程。实验证明,与当前主要MSD(mono、swap、discontinuous)等调序方法相比,该方法较好地体现了维吾尔语的粘着性特点,提高了译文质量。Aimed at the serious problem of phenomenon that there are so many out-of-vocabulary words and translation disor- der in Uyghur-Chinese machine translation, and combined the features of Uyghur and the maximum entropy classification algo- rithm,this paper presented a max entropy reordering model in Uyghur-Chinese machine translation based on tackiness fuzzy rules. The model was based on＂ max entropy classification algorithm, and created the constraints of tackiness fuzzy rules for Uy- ghur in the word level, then extracted better effective reordering rules from training corpus to guide the translation decoding procedure. The experiments show that compared with the current reordering methods, such as MSD （mono, swap, disconti- nuous） and so on, this reordering method is better retained the tackiness feature of Uyghur, and the quality of the translation is also improved.

关 键 词：维汉机器翻译 形态学 粘着性 模糊规则 最大熵 调序模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]