Micro-blog Sentiment Analysis Based On Semantic Sentiment Space Model

Internet has become a major platform for people to share information and exchange ideas inthe past two decades. This is particularly true when twitter and Micro-Blog appear as the newgeneration of social network media platform, which enables people to share information in a free,rapid and flexible manner. Analysis of Micro-Blog information is therefore of vital importance.For instance, the authority can use it to catch public consensus in a timely manner, andbusinesses can get users feedback of their products or services in pursuit of improvements.This paper targets Micro-Blog analysis, in particular the analysis of its sentiment. The key ishow to put forward a method of Micro-Blog orientation classification to resolve the problem ofMicro-Blog sentiment analysis. Based on a detailed study of the various features of Micro-Blog(e.g. links, emoticons, pictures, and labels etc.), and its traditional Chinese semantic features, wedevelop an approach to extract its semantic features through dictionary resources and wordcombinatory units. Then we use dimension-space theory to build a sentiment space model basedon Micro-Blog ten-dimension semantic features. This model can simulate the positive andnegative sentiment states. Finally, combined with Micro-Blog features, we use SVM to designand implement a Micro-Blog sentiment analysis system.Preliminary experimental results show that the approach proposed above provides aneffective solution to Micro-Blog sentiment analysis. The classification accuracy reaches80.6%and79%for positive and negative Micro-Blog respectively. And it validates the feasibility andrationality of this method.