Research On Mongolian Dependency Parsing Based On The Conversion Of Chinese-Mongolian Dependency Parsing Tree

Syntactic analysis is one of the key technologies in Natural Language Processing. Syntactic analysis aims at eliminating the lexical ambiguity and the structural ambiguity in a sentence. Syntactic analysis plays a crucial role in Natural Language Processing. On the one hand, syntactic analysis can effectively verify the correctness and completeness of the treebank; on the other hand, syntactic analysis can directly provide service for the up-application, such as keyword recognition and the visiting log analysis of search engines users. In recent years, the dependency syntactic structure has been a hot spot in the field of Natural Language Processing, which owns to these advantages:simple in form, easy to tag and use. However, there are few researches about the Mongolian dependency parsing.In this thesis, I put forward a new method on the Mongolian dependency parsing based on the conversion of Chinese-Mongolian dependency parsing tree under combining the advantages of phrase structure tree and dependency structure tree. First of all, I use the Stanford parser to analyze Chinese sentence which based on phrase structure, and create the Chinese phrase structure tree. Secondly, according to the word order character of Mongolian and Chinese, I build the model of word order to remain the Chinese word order consistent with the Mongolian word order. Then, I define the transformation rules, using the center node to achieve the conversion, which guides the conversion from the phrase structure to the dependency tree. After pruning process I get the Chinese dependency structure tree. Finally, I create the Mongolian dependency structure tree by taking advantage of the Chinese-Mongolian bilingual word alignment, using the joint probability model, and post-processing the special grammatical phenomenon in Mongolian.Experimental results show that the dependency syntactic analysis of Mongolian combined the characteristics of phrase structure tree and the dependency structure tree can effectively take the advantage of the phrase structure internal characters and phrase external relations to achieve the study of the Mongolian dependency parsing. The accuracy of the Mongolian dependency parsing is 77.31%.