Sentiment Classification in Chinese Microblogs: Lexicon-based and Learning-based Approaches

Sentiment classification in Chinese microblogs is more challenging than that of Twitter for numerous reasons. In this paper, two kinds of approaches are proposed to classify opinionated Chinese- microblog posts: 1) lexicon-based approaches combining Simple Sentiment Word-Count Method with 3 Chinese sentiment lexicons, 2) machine learning models with multiple features. According to our experiment, lexicon-based approaches can yield relatively fine results and machine learning classifiers outperform both the majority baseline and lexicon-based approaches. Among all the machine learning-based approaches, Random Forests works best and the results are satisfactory.