Research On Chinese - Naxi Syntax Machine Translation Based On Tree To Tree

Machine translation is important to communicate between different languages, and it is the hot and difficult area in the field of natural language processing. In recent years, researcher gives more attention to information application of the Naxi minority language all over the world, so it is very important to realize intercommunication between Naxi language and Chinese. In order to have a good translation between Naxi-Chinese machine. it is necessary to introduce the Naxi syntactic information, and there is a lot of lack of alignment problems between Naxi language and Chinese, In this paper, for the purpose of use Naxi syntactic information efficiently, we mainly achieved the following results:(1) For the purpose of use Naxi syntactic information efficiently, we put forward a method of Chinese-Naxi syntactic statistical machine translation based on tree-to-tree. First of all. for effective use of the source language and target language syntax information.we collected Chinese-Naxi aligned parallel corpus and syntactic analysis the corpus, resulting in a corresponding Chinese-Naxi phrase structure trees. Then using GMKH algorithm to extract the translation rules between the fragment of Chinese phrase tree and the fragment of Naxi phrase tree,to get the translation templates according to the probability of a large number of translation rules. Finally.using the tree-parsing algorithm and the translation template to guide the decoding, translation of each source language of Chinese phrase tree fragments in a bottom-up fashion.to obtain the final translation.Comparison with the tree-to-string model, experiments show that this method improves the1.2BLEU values, indicating that the method is effective use of the Naxi syntactic information to improve Chinese-Naxi syntactic statistical machine translation.(2) Based on the alignment of the sub-tree Chinese-Naxi language translation methods.We proposed Naxi language syntax for the characteristics of the sub-tree align and integrate translation model training methods, in order to solve the the alignment of the Naxi language problems. Experiments show that the proposed sub-tree based on the alignment of the tree to tree translation template improves the best selection of translations.(3) We using the alignment tool, combination of the phrase tree, tree translation template, decoding algorithm and the language model to bulid the statement statistical machine translation prototype system...