Research And Application Of Chinese Dependency Parsing Based On Rules And Statistics

Parsing is one of the important subjects of natural language processing. The dependency parsing aims at deriving the syntactic structure of an input sentence automatically according to the dependency grammar. Compared with more informative lexicalized phrase structures, the dependency parsing has the advantage that it provides a simple description of the syntactic relations in a sentence that could be easily understood and converted to semantic dependency description, and dependency parsing has more widely applications, such as machine translation, relation extraction, and ontology construction and so on. This paper investigates the techniques and applications of Chinese dependency parsing based on rules and statistics mainly in the following points:Firstly, the paper proposes an identification method of maximal-length noun phrase based on maximal-length preposition phrase. This novel technique utilizes the mutual restricting characteristic of maximal-length noun phrases and adverbial maximal-length preposition phrases, and uses new tags and above long-distance word as features, which is in favor of the dependency tree building.Secondly, the paper presents a novel method for multi-stage statistical dependency parsing based on dependency direction. In the method, dependency parsing processes are divided into multiple sub-stages, and each stage is in a sequential pattern, which makes it easier to take applicable solutions for different issues in dependency parsing. Meanwhile, dependency parsing in the previous stage provides a clearer context for next stage. Furthermore, due to the dependency direction, the proposed method has lower search complexity than classic graph-based methods.Thirdly, by utilizing the advantages of the rule method and the statistical method, this paper researches on the dependency parsing techniques based on rules and statistics. The statistical part uses the multi-stage statistical dependency parsing method based on dependency direction, and in the rule part, this paper uses the word lists based on statistics. Experimental results show that compared with common methods, the proposed method in this paper offers comparable accuracy and higher efficiency.The finally, for the applications of the dependency parsing, the paper proposes a template building method. This method, based on rules, builds the templates of sentences automatically by using dependency arcs and dependency relations, which provides help for machine translation and information retrieval effectively.The first three divisions consist in the research of Chinese dependency parsing methods in theory. The first three divisions of the paper serve for the last one. The last division of the paper mainly investigates the practical explorations of the Chinese dependency parsing.