Research on sentiment analyzing in multi-topics texts

Abstract

Many web texts coverage multiple topic. How to analyze sentiment this kind of texts is a key problem in information reveal. In this paper, we proposed a new method to analyze sentiment of multi-topics texts on the Internet. Firstly, the Latent Dirichlet Allocation model is used to reveal to the latent topical facets in texts. Then a two a two-layer conditional random fields is introduced to analyze sentiment of every topic in the texts. Empirical results show that this approach is effective for extracting subtopics and identifying sentiments of each topic. Moreover, this method is quite general and can be applied to any kinds of web texts. © 2011 Springer-Verlag Berlin Heidelberg.