Topic Sentiment Change Analysis

Public opinions on a topic may change over time. Topic Sentiment change analysis is a new research problem consisting of two main components: (a) mining opinions on a certain topic, and (b) detect significant changes of sentiment of the opinions on the topic and identify possible reasons causing each such change. In this paper, we discuss topic sentiment change analysis using data on the Web. We adopt probabilistic topic model and language grammar based sentiment analysis techniques, and integrate them together into a topic level sentiment analysis method. This method is capable of analyzing sentiment and identifying sentiment changes of a given topic from a set of documents covering this topic and possibly other topics. In addition, as the contents of relevant topics are differentiated, our method is also able to identify hot events which are possible causes of a sentiment change. Experimental results show that our method is very promising.