Study On Several Issues Of Knowledge Management And Text Mining

In knowledge economy era, how to improve more effectively knowledge management for enterprises is one of research directions of management science. Aiming at above issues, some issues for knowledge management and text mining have been studied substantially as follows:Through analyzing the existing models of knowledge chain and the theory of knowledge pullulating, some new models of knowledge chain and a new perfect mode of knowledge pullulating are put forward. And the DEA method is used to analyze and evaluate the systems of management. These viewpoints provide theoretic foundation and implementing criteria for knowledge management of enterprises.A new method of text clustering based on self-organizing neural network is presented in this dissertation. Traditional text clustering methods, for example, K-means method, need to give the number of clustering seeds, which will impact on the result of clustering. On the contrary, the new method of text clustering can produce automatically a proper new category, which makes up effectively the defect of the K-means method, and the new method can adjust the precision of text clustering through regulating an input parameter.A new method of text clustering and a new method of text categorization based on LSA and Kohonen network are presented respectively in the dissertation. The theory of LSA is applied to the method, which makes the meaning of each dimension in the vector space model (VSM) change greatly. The meaning of each dimension does not denote simple frequency of term, while includes semantic information. The dimension of VSM is decreased greatly, and so the speed and the precision of clustering are advanced.A new method of decreasing text features based on PA and LSA is presented in the dissertation. The method applies the theory of PA to constructing the VSM, and the words having the same contribution to categorization are ascribed to the same pattern, then the dimension of the vector in the VSM is greatly decreased. At the same time the function of seldom words is intensified, which advances the speed and the precision of text categorization. The application of LSA decreases further the dimension of the vector, and embodies semantic information in the VSM, which advances further the speed and the precision of text categorization.