Microblog Sentiment Classification Based on Supervised and Unsupervised Combination

Abstract

Millions of users are sharing their views and opinions in microblog everyday, which makes sentiment classification in microblog be an important and practical issue in social networks. In this paper, we combined the conversional supervised algorithm with unsupervised methods to conduct sentiment analysis. Specifically, we divided the content into two parts: those with emoticons and those without emoticons, and use multiple optimization for the two different parts. Practical evaluation shows that our methods could perform effectively and efficiently for this attracting problem. © Springer-Verlag Berlin Heidelberg 2014.