Chinese Text Feature Dimension Reduction Based on Semantics.

Feature dimension reduction is an important step in text categorization, but traditional feature dimension reduction method ignores semantic information of features. In order to solve this problem, this paper, with the semantic dictionary, proposes a new feature dimensionality reduction processing method. The word-semantic knowledge base is constructed on the basis of HowNet and The Semantic Knowledge-base of Contemporary Chinese. By using the knowledge base and the feature extraction method, text feature is mapped to semantic feature and the dimensional reduction of feature space is realized. Naïve Bayes method is introduced to verify the categorization performance. The experimental results indicate that the proposed approach has a good performance of high dimension reduction and categorization. © 2013 Springer-Verlag.