Dividing for Combination: A Bootstrapping Sentiment Classification Framework for Micro-blogs
Songxian Xie, Ting Wang · 2013
There are many challenges for sentiment classification of user-generated content (UGC) on social media platforms such as micro-blogs. Context dependence, which has been the most challenging problem, is focused on in this paper, and a novel semi-supervised framework is proposed to address the problem. By dividing the feature space of sentiment classification into two parts including the general features and the context features, a general classifier and a context classifier are learned separately in the two partial feature spaces, and a semi-supervised framework is developed to combine the general classifier and context classifier into a bootstrapping classifier. Experimental results show that both the general classifier and context classifier outperform traditional lexicon-based classifier, and the combined bootstrapping classifier outperforms supervised classifier upper bound. The proposed semi-supervised framework is flexible and effective in solving the context dependent problem of sentiment classification for micro-blogs without the need of labeled data.