The Research On Chinese Question Answering Technology Based-On Pattern Matching

Question answering (QA) technology is a very hot research direction in the field of nature language processing (NLP), and it combines a large variety of NLP technologies. So at present, many scientific research institutes are engaged in the research of QA technology. The question answering technology based on pattern matching (QATBPM) is a kind of very effective QA technology, and it has been applied successfully in the TREC.Presently, many research institutes at home and abroad have been investigating English question answering technologies based on pattern matching (EQATBPM), but few institutes are doing research on Chinese question answering technology based on pattern matching (CQATBOM). In this paper, we try to investigate some technologies for CQATBOM. This paper will introduce two following parts of CQATBOM: off-line pattern learning technology and on-line answer recognizing technology.Off-line pattern learning technology is the kernel of CQATBOM, and its tasks are that how to get the patterns of a question type and that how to compute the precision of the patterns at the off-line station. In this paper, we explore the technology of automatically obtaining the patterns with suffix tree. Because the precision of pattern has an important effect on CQATBOM, so we put forward a new pattern precision model which takes account of the answer types, not using the traditional pattern precision model. In this paper, we do some answer recognizing experiment under the old and new pattern precision model. The experiment result shows that a designed result has been achieved with the new pattern precision.On-line answer recognizing technology can accepts and analysis user's nature-language questions and returns the answers to user directly. It has three following parts: question analysis, information retrieval and answer extraction. Question analysis of CQATBOM includes question classification, question word selection, keyword selection and keyword extension. For a question submitted by a user, firstly we clarify the intention of user's question with question templates. Information retrieval aims to retrieval some documentsrelated to the question submitted by the user. Now, information retrieval technology is very mature, so we do it by calling the web search engine. Answer extraction will find the candidate answers from the documents with the patterns, calculate their precision, store them, select some that their reliabilities are bigger than others, and return them to user.At last, this paper introduces the system realization of Chinese question answering technology based-on pattern matching.