| The growth in popularity of the Internet highlights the importance of developing to extract suitable answers to queries targeted at large unstructured corpora. Question Answering (QA) system has become one of main trend for this purpose. It is expected that it can completely or partly take the place of current information retrieval technology based on string match and provide an easier man-machine interface for information acquisition. The questions with natural language form are as input by users, then are analyzed and processed by QA. QA searches candidate answer information from network or other information sources, and evaluates these candidate answer information for generating answers with natural language form.QA's final goal is to extract answers for question sentence from information sources. It needs to know what information will be searched and what condition will be satisfied by possible answers through analyzing the question sentence. So it has become an important task how to analyze and understand question sentence in question sentence processing or even in QA.In this thesis, the characteristics of current question parsing techniques are analyzed. Based on the analysis, combing with some natural language processing (NLP) technology, an approach of question parsing method based on syntactic structure parsing and semantic information is proposed. In the question parsing, the question classification based on modified Bayesian algorithm and query expansion are implemented by applying dependency syntax parsing and semantic web. The related algorithms, processes, arid the key techniques in the question parsing system are introduced in the thesis.The experiments for evaluating question parsing system are designed and implemented. The experimental results show that method proposed in this thesis can get better precision of classification, and the question sentence can be better understood using this method and the retrieval results can be more effective in QA. |
PDF Full Text Request