Using SVM-based LLSI for text classification
使用基于SVM的局部潜在语义索引进行文本分类

Latent Semantic Indexing (LSI) uses Singular Value Decomposition (SVD) to obtain latent semantic structure of original term-document matrix, and problems of polysemy and synonymy can be dealt with to some extent. However, the present available methods of applying LSI to text classification are not satisfying, since they do not take full account of classification information. To solve the problem, an improved Local LSI (LLSI) method was proposed, using Support Vector Machine (SVM) to produce the local region. Experimental results suggest that the proposed method is effective.