Multi-class Sentiment Analysis

Abstract

The popularity of the web and mobile devices, people can generate messages anytime from there to express their opinion and emotions. The job of sentiment analysis is to analyze the emotional state of the person who left the message. Sentiment analysis can be two-class or multi-class, in this paper, we propose a method of extracting sentiment words for multi-¬class sentiment analysis and use the Ministry of Education Dictionary as supplement to expand the sentiment words according to the extracted sentiment words in the training dataset. The experimental results show that the accuracy of our approach is close to the best method and higher than the other methods for multi-class sentiment analysis.