Research And Design Of Restricted-Domain Question Answering System

With the development and application of the Internet, on-line information is growing rapidly. It hopes that we can obtain the information needed from the mass of the network content. The search engine has largely solved this problem. People only need to enter some keywords, the search engine will return relevant pages. However, faced with so enormous information in pages, it is difficult for users to find what they need quickly. Therefore, in order to meet the desire that people are able to obtain information more quickly and accurately, automatic question answering (QA) system gradually developed.Automatic QA system allows users to use natural language to ask questions, and returns a concise and accurate answer to the problem. It uses a variety of natural language processing (NLP) techniques and is one of the hot spots in computer application domain research. At present, the English QA system is more mature, but because of the complexity of the Chinese natural language, the Chinese question QA system is still at a preliminary stage. This paper researches a Chinese QA system in the restricted domain.In this paper, according to the characteristics of computer domain knowledge, we design a Chinese QA system based on frequently-asked question (FAQ) in the computer network course. The system focuses on building domain knowledge base, problem understanding, computing sentence similarity algorithm and other aspects. In the part of building the domain knowledge base, we research and design the structure of course points table, FAQ storage, FAQ pretreatment. In the part of problem understanding, we mainly study the Chinese segmentation, the keyword extraction and expansion, the problem of classification and so on. In the part of sentence similarity computing, we use the similarity calculation algorithm based on semantic similarity. And we build the corresponding test set for the pilot test. In the last of the article, we describe the experimental results and evaluation of the entire automatic QA system.