Based On The Sequence Annotation Of Chinese Dependency Syntax

More and more research in the field of syntactic analysis has been focused on dependency syntactic analysis in recent years. Compared with the phrase-structure grammar, dependency grammar has the advantages of concision form, semantic oriented, and analyzing easily. Therefore, dependency analysis has been widely applied in many practice fields, such as document summarization, open domain answering question system, machine translation.This paper focused on dependency analysis. First, this paper gives an detail introduction of related works, including the basic theory of dependency grammar, previous dependency analysis methods, and recent discriminative dependency syntactic models. Then, a new dependency analysis method based on sequence labeling model was presented in this article. In our method, we divide the problem into 3 phases. In the first phase, we need to change dependency structures into sequences; then we build a dependency sequence model via the sequence labeling model in the second phase; in the last phase, we use the model to predict the dependency structure. We focus on the first and the last phases in this paper. The first phase is the transforming phase. It always requires that the transformation should keep the original information of dependency structures, as well as benefit the building model phase. We propose a transforming method. According to our method, the whole structure information are been kept. Moreover, we improve the decoding algorithm in order to make it suitable for the new task and boost the performance. Besides, the experiments proved that our method could adapt the dependency analysis task.At last, the research achievements are concluded and an affirmative outlook is presented for future works in the new dependency analysis framework.