| Question Answering System (QAS) is the next generation of search engine. It allows the user to ask questions using natural language and returns precise answers. So, compared with traditional search engines, QAS can retrieve the most precise answer to satisfy user's demand of searching. Natural Language Processing (NLP) is used to research the two key techniques which are question analysis and information retrieval in this paper and realized a Chinese QAS named CQAS.Question analysis is the initial task of Chinese QAS, the result of which has a great effect on following processing work. we make the following research based on the previous studies: (1) The rules combined the question word with question focus based on heuristic is proposed to classify questions due to the system mainly deals with simple factoid questions; (2) Question sentence pattern database is built, sentence pattern analysis is carried out, and that improves the capability of processing the questions in different formats; (3) Semantic role labeling is done using the machine learning method and maximum entropy model as classifier, and changes the questions into semantic frame.Information retrieval is a very important connecting part in QAS, which directly effects QAS performance. We perform a two-strategy retrieval method. It includes the local knowledge database retrieval and web retrieval. The local knowledge database is established from knowledge sharing platform to do local retrieval, which reflects the users' real problems. If local knowledge database retrieval failed, then web retrieval is done, the web search engine Google is used to do web retrieval, QAS only uses the snippets of search results instead of downloading the network source documents.Finally, we do the tests and analysis of the technique applied to question analysis and information retrieval in our system CQAS. The experimental results indicate the rationality of the theories in this paper. |
PDF Full Text Request