Textual analysis for online reviews: A polymerization topic sentiment model

Abstract

More and more e-commerce companies realize the importance of analyzing the online reviews of their products. It is believed that online review has a significant impact on the shaping product brand and sales promotion. In this paper, we proposed a polymerization topic sentiment model (PTSM) to conduct textual analysis for online reviews. We applied this model to extract and filter the sentiment information from online reviews. Through integrating this model with machine learning methods, the results showed that the prediction accuracy had improved. Also, the experimental results showed that filtering sentiment topics hidden in the reviews are more important in influencing sales prediction, and the PTSM is more precise than other methods. The findings of this paper contribute to the knowledge that filtering the sentiment topics of online reviews could improve the prediction accuracy. Also, it could be applied by e-commerce practitioners as a new technique to conduct analyses of online reviews.