Research On Chinese-uyghur Word-alignment For Statistical Machine Translation

In the field of natural language processing, such as the constructing of corpus, speech recognition, bilingual dictionary compilation, word-alignment provide fundamental construction. At present, Chinese-English word alignment technology research has achieved accuracy rate at 90.0%, the recall rate at 88.2%, however Chinese-Uyghur word-alignment research start lately. This paper mainly studied Chinese-Uyghur word-alignment in the sentence level, using approach of statistical machine translation.This paper described a Chinese-Uyghur word-alignment system on IBM Model1-3 and Heuristic optimization algorithm.System is divided into two modules: preview process and word alignment process. Word alignment process is: first, use IBM model1-3 to realize Chinese-Uyghur one-to-one and one-to-many and so on, and then use the heuristic to optimize theory of Och to design Chinese-Uyghur many-to-one and many-to-many. Experimental results show that the method is feasible, and reached the entry-level stage expected. The research provides a powerful platform in future.