Chinese sentiment analysis for commodity price level fluctuation news comments
Yan Zhao1, Suyu Dong1, Jing Yang2
1College of Management, Inner Mongolia University of Technology, Huhhot, China
2College of Mechanics, Inner Mongolia University of Technology, Huhhot, China
With the rapid development of the Internet technology and news media, people pay more attention on news especially about commodity price fluctuation. Hence, more and more Chinese news comments about commodity price fluctuation appear on Internet. These comments contain all kinds of sentiment. Analysing the sentiment of these comments will make government know more about Netizen emotion on this information and enhance efficiency of management, which has important practical significance. In this paper, we adopt three supervised learning methods (naive Bayes, maximum entropy and support vector machines) to automatically classify user comments as two classes (positive and negative). Through a lot of experiments, we found that machine learning techniques perform quite well in the domain of commodity price fluctuation news comments sentiment classification. Meanwhile, the effects of the feature representations and dimensions for the classification of the three machine learning techniques are analysed and discussed in detail. Experimental results show that maximum entropy classifier is best overall. Frequency is a better method of feature representation, which can use fewer features to get better result.