Question Dependency Syntax And Semantic Analysis Research

Question analysis is the basic work of question-answering system. The effect of question analysis has an important influence on text retrieval and answer extraction in question-answering system. Question dependency syntax and semantic analysis analyze the dependency structure and semantic components of questions from the dependency syntactic conversion, and it is a mainstay for question syntax and semantic understanding. The thesis focuses on the key technologies of question analysis process, such as question dependency parsing, question semantic components labeling, question similarity calculation to do the discussion and analysis. The main research work completed the following characteristics:1. For the problems of question trunks and focus extraction by conventional dependency syntax parsing, this thesis presents a method of question trunks and focus extraction oriented to question dependency syntax conversion with the characteristics of the questions. The method defines a number of conversional rules of question dependency relations. According to the rules, we conduct merger, conversion and removal for parts component of question syntax, and extract the question trunks, question focus, and other structure information based on question dependency syntactic structure. The experimental results show that the proposed method of question trunks and focus extraction based on dependency syntax conversion achieved good results.2. This thesis realizes a method of Chinese question semantic components recognition. The method defines 30 semantic components of questions at first, and then defines the features of semantic components identification for the characteristics of questions. We label a number of training data on the basis of question dependency syntax structure, and train them with CRF (Conditional Random Field) learning algorithm, and achieved a recognition model of question semantic components. The experimental results show that the question semantic component identification achieved good results.3. The thesis presents a method of question similarity calculation integrating with dependency syntax and semantic components. According to the question dependency syntactic structure, this method extracts dependent pair sequences based on question dependency relationship at first, and then identify semantic components, finally compute the question similarity by calculating the similarity of dependent pair sequence and semantic components. The comparative experimental results show that the question similarity calculation method of integrating dependency syntax and semantic analysis achieves good results for taking into account the syntax structure and semantic information.4. We build a prototype system of question dependency parsing and semantic components identification with the above results, and design and implement a question similarity calculation prototype system of integrating dependency syntax and semantic analysis.