This paper introduced a clustering-based Chinese sentiment analysis approach which is a new method to sentiment analysis appropriated for short text such as Sina Weibo. By building Sentiment Sequence from Weibo text, we apply the Longest Common Sequence algorithms to measure the sentiment different from two Sentiment Sequence, and K-Medoids clustering method to break a mass of Sentiment Sequences into groups. It has great advantages over the existing sentiment analysis method such as classification by supervised learning algorithms. The experiment result shows the sentiment distribution and groups of a mass of Weibo data aggregated by a given topic, and in a specific period. The method is well performed, efficient, and non-human participating, and appropriated for Chinese short text.
Wei, Guannan; An, Haizhong; Dong, Tiancheng; and Li, Huajiao, "A NOVEL MICRO-BLOG SENTIMENT ANALYSIS APPROACH BY LONGEST COMMON SEQUENCE AND K-MEDOIDS" (2014). PACIS 2014 Proceedings. 38.