The Construction And Application Of Patent Term Knowledge Base

Patent is an important part of scientific and technical literature, and it is playing a more and more important role in scientific research and technological innovation. In patents, technical terms are the core content and an important component. The analysis and research on terms are the basis of the effective application of patents.Term knowledge base is a semantic system which is used to store and describe term concepts and term relations. Focusing on the construction of a term knowledge base and its application in a patent retrieval system, this paper presented the related algorithms of the extraction of term relation and the calculation of term similarity at the aspect of algorithms, and implements a query expansion model based on the term knowledge base at the aspect of system construction. The main work of this paper includes:To solve the difficulty of the automatic construction of a knowledge base, namely the automatic acquisition of high-quality knowledge, this paper presented the extraction method of relation features based on a latent relation analysis. In this method, firstly, the features of term pairs are extracted from patents, and the features are extended by syntactic information; then singular-value decomposition is performed to the relation feature matrix by a latent relation analysis, and the latent data relevancy between feature vectors is analyzed, so as to obtain a feature vector matrix; and finally, relation types are identified with SVM.The calculation of term similarity is the core issue of query expansion. However, the noise relation in term knowledge base tends to produce erroneous expansion words. To solve the problem, this paper presented a context-based term similarity calculation method. This method uses the membership function in fuzzy mathematics to calculate the similarity of relevant information in the context of terms in order to make up the deficiency of query expansion method only using a term knowledge base.Taking the term relations in the term knowledge base as a query expansion resource and the term similarities as expansion constraints, the queries of retrieval words are expanded. Experimental results on patent retrieval show that compared to the query expansion method based on a global analysis, the new query expansion method leads to higher F-values of the returning result of the retrieval system.