Affective-word based Chinese text sentiment classification

When browsing news on the web, various emotions may be evoked in readers and furthermore cause different influence on their minds and life. We expect that emotional analysis and classification of text may provide good performance and significance to users surfing the Internet. Most previous research only focus on bi-emotion classification, that is, Positive and Negative, e.g., identifying whether a comment is for praising or criticizing. In this paper, we propose a χ²-based Chinese text emotion classification with five sentiment categories. We run two experiments, one uses sentiment words extracted from HowNet and a Chinese thesaurus: TongYiCi CiLin, and the other is not. The results shows that adding affective words can make better prediction in the sentiment classification.