A novel Naxi dependency parsing method based on rules and statistics

This paper presents a Naxi dependency parsing method which combined rules and statistics, based on the characteristic of the Naxi language. Firstly, the annotation standard of Naxi Dependency Treebank was established and dependencies were defined which are based on the syntactic structure characteristics of Naxi language, then Naxi Dependency Treebank was constructed; Secondly, the constructed Naxi dependency Treebank was used as the basis, and the Naxi language phrases were analyzed independently by means of rules, then the boundaries and categories of phrases were determined. The features of core words postposition in the Naxi language were regarded as a condition of constraint. Then the dependencies between phrases were analyzed. Finally, the dependencies between phrases were combined, and interdependent probabilistic models were used to parse the syntax of Naxi language, and finally a completed Naxi language dependency parsing system was achieved. The results of dependency parsing comparative experiments show that the use of Naxi dependency rules improves the performance of Naxi language dependency parsing system. Moreover, the performance of the proposed method is better than the dependency parsing methods which are fully based on statistics.