Research On Chinese-Chinese Bilingual Ordering Model Based On Syntactic Rules

In recent years, information processing technology research of ethnic minorities is more and more getting people’s attention. To informationize of ethnic minority language by using natural language processing not only helps to improve the information level of the ethnic minorities, but also can strengthen information communication and fusion between nationalities. Reordering problem in machine translation is caused by the non-monotonic mapping between languages, lead to decline in the quality of the translation. This non-monotonic feature exists between Yi languages and Chinese. Therefore, the reordering problem becomes one of the key researches in Yi-Chinses machine translation.This thesis proposed a solution to reordering in the phrase structure tree of Yi. Firstly, Yi parsing is carried out according to the features of Yi language; secondly, reordering rules are obtained by comparing Yi and Chinese; and finally, a reordering model is generated based on phrase structure tree of the source language and the bilingual word alignment result. Under the guidance of reordering rules, the reordering model is able to reorder the string of source language to conform the logical structure of target language. The main work of this thesis is as follows:Realize the parsing for Yi: before contrasting the structural differences of Yi and Chinese, this thesis study syntactic rules of Yi, and realize the parsing for Yi language. The result of paring is the phrase structure tree of Yi as the input of the reordering model.Obtain the reordering rules: this thesis proposed a series of reordering rules, based on the structural differences between Yi and Chinese, to guide the reordering operation in the model. Furthermore, in this paper, we take experiments to verify the correctness of the reordering rules; the experimental result shows that the rules have positive effect on reordered result.Propose reordering model: this thesis proposed a reordering model to reorder the phrase structure tree of Yi, get source language strings that conform the logical structure of target language. In addition, we take reordering experiments to test the model’s effect on improving the quality of the translation. The result shows that the score of BLEU after reordering is superior to before.