Lexicon-Based Sentiment Analysis on Topical Chinese Microblog Messages

Microblogging is a popular social media where people express their opinions and sentiment on social topics. The Chinese microblogging service, called Weibo, has become a remarkable media in the Chinese society. People are eager to know others’ attitudes towards social events, thus sentiment analysis on those topical microblog messages is important. In this paper we introduce a lexicon-based sentiment analysis method. We construct a Weibo Lexicon with representative topical words and out-of-vocabulary (OOV) words, which are usually informal and are not existing in formal dictionaries. In addition, we use a propagation algorithm to automatically assign sentiment polarity scores to the words discovered. These scores are more closely reflecting the Weibo context since words may have new or opposite polarities instead or their formal meanings. Evaluations on the classification tasks show that our method is effective on recognizing the subjectivity and sentiment of Weibo sentences. The Weibo lexicon increases the performance of the classifications.