Detecting Microblogger's Attitude towards Bursty Events: a Text Chain Model

摘要

With the booming of social media, microblog attracts more and more people to discuss public issues and share their views and opinions. In this paper, we focus on the sentiment analysis in Chinese microblog from the aspect of users. We aim to detect microbloger's attitude on bursty events by proposing a novel text chain model. We firstly formulate the problem of user sentiment analysis. By leveraging the link symbols in contents, we generate micrblog units and prune to user text chains which will be regarded as a whole in the follow-up process. Then, we use MaxEnt-LDA model to extract target events and opinion words, and use a lexicon-based model to detect user's orientation towards a certain atomistic event. Experimental results show that our model could detect user's attitudes effectively.