关键词: 知识库, 特征提取, 粗集, 属性约简

Abstract: At present，field features in many knowledge bases mainly rely on experts by hand to obtain，this process is not only time-consuming and laborious，but also the quality of knowledge base is restricted by knowledge and experience of experts.Although there are some methods of automatically obtaining field features，but obtained feature sets are redundant and not representative.So it is an urgent problem that how to design the method of automatically obtaining representative and low redundant field feature subsets.The thesis firstly analyzes some methods of automatically obtaining field features and finds out their shortcomings，and then presents an improved feature extraction method.Finally，combining the improved method with rough sets to automatically obtain lower redundant and higher representative field feature subsets of large-scale corpus.Experimental results show that the comprehensive algorithm is efficient and practical.

Key words: knowledge base, feature extraction, rough set, attribute reduction