Based on HowNet and the Semantic Knowledge-Base of Modern Chinese (SKBMC),the Word-Semantic Knowledge Base (WSKB) was constructed.By using the WSKB and the feature extraction method,text feature was mapped to semantic feature and realized the dimensional reduction of feature space.Na(i)ve Bayes method was introduced to verify the classification performance.Experiment result shows that classification performance of individual class increased own to excessive generalization of the feature space,so the overall classification accuracy did not changed significantly.