Comparison of SVM classification method and semantic similarity method for sentiment classification

With the growth of the Internet and electronic commerce, there is more and more review data on the Internet. Quite a lot of Internet users refer to related comments of a product before they make a decision, which can teach them about the quality and reputation of the product and help them decide whether to buy it. A system that can automatically classify the polarity of a given text would be a great help to users. This paper is divided into two parts. The first part is about SVM classification method. We adopts a variety of feature extraction methods, such as TF-IDF (Term Frequency-Inverse Document Frequency), MI (Mutual Information), CHI. First, we calculate the weight of terms. And then, we adopt the Support Vector Machines (SVM) model for emotion classification. In the second part, we present a new method of semantic similarity computation for sentiment analysis. Our approach achieves the accuracy of 91% and 82% with the two methods.