Question Analysis And Answer Extraction Of Chinese Question Answering System

In the era of rapid development of today's information, the method to get something useful rapidly and effectively from the flood of information shows great significance. Question answering system can saves time and improves efficiency for users that supports the natural language input and returns a precise answer. It consists of question analysis, information retrieval and answer extraction. The question analysis module and the answer extraction module were chosen in this paper. Based on the domestic and foreign research results, the research was fully studied in terms of feature extraction, question classification and similarity calculation.The machine learning method was used for the research of question classification. The selection of features and classifier palyed important role in the accuracy of question classification. In this work, we experimentally verify a new method that the classifier support vector machine. First, the question sentence was pretreated to extract the basic characteristics, thereby resulting in a new set of features. Then the new features was encoded in a format that can be recognized by the classifier. Based on the proposed strategy, the method showed a satisfactory classification performance.In the study of answer extraction, there was a problem that the syntactic and semantic features of the sentence can not be taken full account. A similarity calculation method was proposed based on dependency tree and dependency phrase tree. The syntactic parsing technology was used to obtain dependency tree and dependency phrase tree. And the convolution tree kernel was applied in similarity calculation. Compared with other methods, the established method showed its excellence.In a word, a prototype of Chinese question answering system prototype was established. This work also provides a good reference for the system in application of education.