Study And Construction Of A Chinese Uyghur Aligned Template Library For Generalized Example Based Machine Translation

Machine Translation (MT) is a technology which researches on how to translate alanguage into another automatically by computer. Among different strategies in machinetranslation, the example-based translation method is accurate, fast and easy to acquiretranslation knowledge. But it will be inefficient when searching on enormous scale corpus inactual applications. So a generalized example-based method is proposed to extract translationtemplates from bilingual corpus to reduce the space dimension of the corpus and to improvethe translation quality and the efficiency in matching rate. This paper studies how to extractthe Chinese-Uygur templates, establish template-aligned library, and use the library fortemplate matching.First, this paper introduced the history of machine translation and the typical strategies ofmachine translation from two aspects. It introduced the current situation of Uyghurinformation construction and Chinese-Uyghur machine translation and showed the necessityand feasibility of template extraction. After detailed analysis of the generalizedexample-based method, a method which could extract Chinese-Uyghur aligned templateautomatically as well as the process of translation by using the template library was showed.Specifically, this paper made tentative works in the following aspects: designed filter toscreen corpus, summarized the common points of the corpus, used lexical as the constraint oftemplate slot; achieved the alignment of words in Chinese-Uygur sentence pair and made thesentences in the corpus to be word aligned; defined the template, realized the Chinese-Uygurlanguage template extraction and successfully set up a small template library; developedChinese sentence template matching algorithm and got Uygur output correctly; verified thefeasibility of the generalized example-based method in Chinese-Uygur machine translation.