Chinese text emotion classification based on emotion dictionary

Recently, much work have been done on text emotion classification. However, they mainly focused on the emotions expressed by authors instead of the readers. In addition, researches on simplified Chinese text emotion classification are extremely less. In this paper, we proposed a simplified Chinese text emotion classification based on readers´ emotions. Mass of documents with readers´ emotion tag are used as raw text sets, and Vector Space Model is used to represent each document. An emotion dictionary is created semi-automatically by using WordNet to build text vectors. We then train a Support Vector Machine classifier on preprocessed data with four emotion classes, and compared the predicate results with that from Naive Bayes classifier. Experiment results indicate that our approach performs much better on classify accuracy and efficiency.