Restricted Domain Of Chinese Question Answering System Questions

Question answering system, as the new generation of intelligent search engine, allows users to ask questions by means of natural language, and can supplly more accurate answers compared with traditional search engines. Question analysis is a very important component of question answering system. The accuracy of question analysis directly affects the accuracy of the ultimate answer extraction. In the paper, Yunnan Tourism FAQ question answering system model is constructed based on natural language processing technology, domain knowledge base construction technology, question formal expression method, question classification and questions similarity calculation.Main distinctive achievements are as follows:(1) A domain knowledge expression, domain ontology extraction and construction method is proposed according to the deficiency of commonsense knowledge base, HowNet, in domain problem describing. The method is ontology oriented and can construct domain knowledge base, integrate domain knowledge base and common knowledge base on description of domain concept provided by HowNet's concept description language.(2) A question formal expression method is proposed. The method realized key words extraction and domain question expansion by lexical and semantic analysis. Syntactic interdependence tree of the question is extracted by question syntactic analysis. Question type, question focus and answer type of the question is obtained by the mapping rules of question type and answer type.(3)A domain question classification method based on language rule and statistical learning is put forward. First, question classification rules are extracted by language rules and domain knowledge. Then, the question classification model is constructed through extracting syntactic structure relation, domain features and improved Bayes classification learning algorithm. At last, domain question classification is realized by combining language rules and statistical learning. Experiment shows the proposed method is feasible.(4)According to the deficiency of current question similarity method, a domain question similarity calculation method combined with the feature of domain Chinese question is put forward. The method calculates the Semantic similarity between words, extracts question syntactic interdependence pairs, and calculates the similarity between question syntactic interdependence pairs based on domain knowledge base and common knowledge base to calculate domain question similarity which combines lexical, syntactic, semantic and domain knowledge. Experiment result shows great efficiency.(5)Collect domain features and implement the Yunnan tourism FAQ question answering system based on the research above.