Q & A Web Forum Of Knowledge-based Automatic Question Answering System

With the development of information retrieval technology, various types of search services have appeared on the Internet. In all the services, the one that is most widely used is Web search engine, which has realized the acquisition, processing, storage and access on the mass of Web documents, in order that users can find necessary information on the Internet quickly and easily. Thus the Web search engine plays an important role in people's daily life. However, for the growth of Internet information and the maturity of search technology, Internet users have no longer satisfied with a keyword-based search service, and hope that they can express their query needs through natural language description, and the search service system can understand their intention to return appropriate results. Therefore, the automatic question answering system has become the next desire of Internet users. Automatic question answering system offers a user interface similar to search engines; while users will no longer commit keywords, but natural language questions. The system will return a list of answers ranked by their association with the question.By present, a great deal of research results has been accumulated in the field of automatic question answering, including those based on different languages and different data sets. However, there is no automatic question answering service yet, as a product as search engine.Aiming at the accumulation of experience for the product of automatic question answering Internet service, this paper studies the construction process of automatic question answering system, which is based on a specific data set - Question and Answer Web Forum data sets. This paper also studies a series of programs to improve the system performance, through the introduction of natural language syntactic and semantic information and new organizational form of data sets into information retrieval process. The main contributions and innovations include:·The construction of evaluation platform for automatic question answering system. Build an automatic question answering system on the question and answer forum data sets, referring to the realization of the search engine, using the term-based text similarity model. Experiments show that: the performance of the system is slightly better than that of their own similar-question-search service in question and answer forums. Then it will be treated as the evaluation baseline of system performance improvement in this paper, instead of the similar-question-search service.·Performance improvement of automatic question answering system based on dependency term. A definition of dependency term is proposed, based on term, integrating the natural language dependency structure, introducing syntactic information into the information retrieval process. Experiments show that: dependency term can effectively express the characteristics of natural language questions, and improves the performance of automatic question answering system without changing of original information retrieval models.·Performance improvement of automatic question answering system based on question classification. A new definition of taxonomy for natural language questions is proposed, for the the question and answer forum data sets, and a question classifier is trained by natural language syntactic and semantic features, which is accurate enough to guide the answer ranking. Experiments show that: the question classifier works well on the Web forum data sets, and the question-class-guided ranking significantly improves the system performance.·Performance improvement of automatic question answering system based on natural language-based knowledge system. Madding references to the former two performance improving methods of the system, a new data organizing form is proposed: adding predicate links to concept system to establish a natural language-based knowledge system. The predicate links are generated from the question/answer pairs in the date sets. This is a comprehensive improvement; making full use of not only the information from answers, but also the relationships form concept system, enhancing the logical reasoning ability of the system. Forum data are all filled into the knowledge system, based on which the automatic question answering system is rebuilt. Experiments show that: the reconstruction brings further improvement to the system performance.