Rating Supervised Latent Topic Model for Aspect Discovery and Sentiment Classification in On-Line Review Mining

摘要

Topic models have been used for unsupervised joint aspect (or attribute) discovery and sentiment classification in on-line review mining. However in existing methods the straightforward relations between ratings, aspect importance weights and sentiments in reviews are not explicitly exploited. In this paper we propose Rating Supervised Latent Topic Model (RS-LTM) that incorporates these relations into the framework of LDA to fulfill the task. We test the proposed model on a review set crawled from Amazon.com. The preliminary experiment results show that the proposed model outperforms state-of-the-art models by a considerable margin.