Fusion Of Syntax And Predicate - Related Chinese - Naxi Machine Translation Studies

Dongba is created and used by Naxi ancestors in Lijiang, Yunnan province. More than500,000people are still in use. It is famous of "the world’s only living ancient hieroglyphics." Naxi pictograph is not only a valuable asset treasure of Chinese civilization, but also the history of the development of world culture. It has been widely concerned around the world researchers. But now, Naxi pictograph is rapidly variability and dead in influence with the various foreign culture penetration and modern civilization. Therefore, it is currently an important research direction of the Naxi ethnic minority languages that how to use this precious resource of information achieves Naxi pictograph resource protection, universal and heritage and promotes the intercommunication and translation each other. It has more high academic research and practical application. Therefore, this paper combines with the existing Naxi-Chinese bilingual dictionaries and Naxi-Chinese bilingual corpus and combines with Naxi characteristic constructs Naxi segmentation tool based on conditional random fields, Naxi-Chinese bilingual word alignment method based on entity constraint and hierarchical phrase Chinese-Naxi machine translation method based on syntactic and semantic. The specific research results are as follows:(1) Naxi segmentation method based on conditional random fields. Like with Chinese, Naxi language also exist the problem of segmentation. Due to using the word in Naxi sentence composes words, we need to take advantage of the characteristics of the Naxi language. The characteristic can be used to position information of the Naxi word to reflect in Naxi sentence. Therefore, the problem that using the characteristics of the Naxi segments for Naxi language sentence is actually the problem that using the word position feature recognizes word entity in Naxi sentence. Because the conditional random field model does not require very strict independence assumption, it can integrate rich features, so it is widely used in segmentation and extraction of entities and has a high accuracy rate. Therefore, Naxi sentence is segmented by conditional random fields.(2) Naxi-Chinese bilingual word alignment method based on entity constraint. There are alignment relationships among entities in the process of bilingual Alignment. Now, bilingual entities can be effectively identified. Therefore, we can use bilingual alignment relationship to improve the accuracy of the bilingual word alignment.(3) Chinese-Naxi machine translation method based on syntactic and predicate association. Now, statistical machine translation method based on syntactic has been currently mainstream. Therefore, this issue analyses deeply the Naxi and Chinese sentence structure and study how to combine Chinese-Naxi language syntactic features and semantic information. On the basis of common characteristics, we integrate syntactic and predicate association on the basis of machine translation model based on hierarchical phrase for the for Chinese-Naxi syntactic characteristic. The last, we complete construction of Chinese-Naxi statistical translation model.(4) Chinese-Naxi machine translation method based on syntactic and predicate association system implementation. Text-oriented and or paragraph Chinese-Naxi translation still exist many difficulties, but for sentence-level Chinese-Naxi bilingual translation, due to be limited sentence level operational level, the implication of the syntactic or predicate association is relatively simple, to a certain extent, it reduce the difficulty of language comprehension. Therefore, the paper will study the Chinese-Naxi bilingual sentence level translation prototype system. According to the sentence grammatical features of Naxi language, we analyze structure of the Chinese-Naxi bilingual translation system and achieve Chinese-Naxi bilingual translation prototype system.