Location:	Meeting Room 1, Ground Floor, JLB
Date:	3 December 2009
Time:	12:30pm - 13:45pm
Speaker(s):	Yulan He - KMi

Sentiment analysis or opinion mining aims to use automated tools to detect subjective information such as opinions, attitudes, and feelings expressed in text. We recently proposed a novel probabilistic modeling framework based on Latent Dirichlet Allocation (LDA), called joint sentiment/topic model (JST), which detects sentiment and topic simultaneously from text. Unlike other machine learning approaches to sentiment classification which often require labeled corpora for classifier training, the proposed JST model is fully unsupervised.

With minimum lexical prior information incorporated, the JST model achieved comparable performance compared to other supervised approaches on the movie review data. We also studied a variant of the JST model, called Reverse-JST, by reversing the sequence of sentiment and topic generation in the modelling process. Extensive experiments show that Reverse-JST performs consistently worse than JST which justifies the original proposal of the JST model. Moreover, unlike other supervised approaches to sentiment classification which often fail to produce satisfactory performance when shifting to other domains, the unsupervised nature of the JST model makes it highly portable to other domains. This is verified by the experimental results on datasets from four different domains where the JST model even outperformed the in-domain supervised approach in some of the datasets. We hypothesize that the JST model can readily meet the demand of large-scale sentiment analysis from the Web in an open-ended fashion.

Save to your Calendar