Domain Knowledge Base Construction Method And Its Application

Knowledge Base plays a very important role in the research of the intelligent information processing. It is the basic resource of Question Answering System, Information Query, Information Acquire and Machine Translation. For some special domains, the Domain Knowledge Base is more important and it is the foundation of how to solve the domain questions. However, how to construct a Domain Knowledge Base is still a difficult work now. In this paper, author's research is based on how to construct the Domain Knowledge Base, including the description of the Domain Knowledge Base and refining domain sememe, marking the Domain Knowledge Base and the similarity calculation of some concept and phrase in the Domain Knowledge Base in a restricted Yunnan Tourism domain. The main innovative achievements are as follows:(1) We realized the database description of the data in Hownet 2008, and imported the sememe and concepts of Hownet into the database. So we can easily acquire relationships of Hownet sememe, Hownet concepts and relationship of Hownet concepts.(2) In this paper, author proposes a method of constructing Domain Knowledge Base. With the help of ontology ideas, the method analyzes terms of domain and refines domain sememe, and then the domain can be accurately described by the KDML language and domain sememe, and finally constructs the Domain Knowledge Base.2002 Terms of tourism domain were accurately described in Yunnan Tourism Domain. The Tourism Domain Hownet was constructed. Above all, it realizes the amalgamation of Hownet and Tourism Domain Hownet.(3) The similarity calculation of concepts and glossaries based on 2008 Hownet is realized in this paper. Meanwhile, we presented an improved way to calculate the similarity calculation in domain by using extended Domain Hownet. With this method, the similarity calculation of the terms in Restricted Domain is effectively resolved.(4) Through the extended Domain Hownet, the similarity calculation of phrase is also realized in this paper. Meanwhile, with the improved similarity calculation method, the similarity calculation of domain phrase can be effectively resolved.