A Text Sentimental Analysis Method Based on Dimension Reduction of CHI Multi-gram Features Mixture

To address the problem of increasing computation caused by high-dimensional features, we propose a method for text sentimental analysis based on dimension reduction of Chi-square statistic (CHI) multi-grams mixture in this paper. It can not only effectively improve the effect of feature extraction, but also precisely determine the feature dimensions, which is different from the traditional methods using experience value. Experimental results show that the proposed method outperforms the exiting methods and the highest accuracy rate reached 94.85%. Moreover, it is proved that our method is universal for the subjective and objective classification as well as the different length of text classification reviews.