Chinese blog clustering by hidden sentiment factors

Abstract

In the Web age, blogs have become the major platform for people to express their opinions and sentiments. The traditional blog clustering methods usually group blogs by keywords, stories or timelines, which do not consider opinions and emotions expressed in the articles. In this paper, a novel method based on Probabilistic Latent Semantic Analysis (PLSA) is presented to model the hidden emotion factors and an emotion-oriented clustering approach is proposed according to the sentiment similarities between Chinese blogs. Extensive experiments were conducted on real world blog datasets with different topics and the results show that our approach can cluster Chinese blogs into sentiment coherent groups to allow for better organization and easy navigation. © 2009 Springer.