Research On The Questions Classification In The Chinese Question Answering System

Chinese Question Answering System, which is a research focusing areas in data mining, is an intelligent information processing system which can understand the issues described in natural language,locate the correct position of the answer and synthesize answer from the mass information. Question classification, as a key step in the process of Question Answering System, is used to classify the question described in natural language, and to try to collect sufficient information related to the question for improving the accuracy and efficiency of the following process steps in Question Answering System.Main work of this dissertation is to study question classification in the Chinese Question Answering system, and try to find a new classification method making use of the characters of Chinese and Chinese questions to improve the classification accuracy of the Chinese questions, and at last to improve the performance of the Chinese Question Answering System.This research provides some new insights into Chinese question classification:(1) On the basis of the investigation on the relation between key words, such as question words and headwords, and question's category,for extracting question words and headwords more accurate, two methods based on part-of-speech-fixing, which are used to improve the performance of the existing method based on part-of-speech and heuristic rules, are given. And a question classification method combining rules and Bayes, which can overcome the faults that rule's coverage is not high, is designed too. Empirical results show the validity of the method.(2) To solve the shortcomings of headwords extraction in previous method, the application of syntactic structure information in classifying Chinese questions is studied, and a headwords extraction method using syntactic structure is designed and some extraction rules are set up at the same time. Empirical results show the validity of the method.