Chinese text classification method based on NMF

Text document classification based on the semantic level is a hot issue in text processing presently. In this paper, a method based on NMF for Chinese text classification is presented. According to NMF, the term-document matrix is decomposed to capture the relation between terms. This method settled effectively the problems of synonym and polysemy. It experimentally shows that, compared with LSI based on SVD, this method has advantages of faster computing speed, less memory occupancy and improvement of classification precision when the dimension reduces markedly.