Sentiment Analysis of Product Reviews Based on JST Model

Abstract

Product reviews are information that users comment after purchasing products online, and it contains user's sentiment information about the product. Considering that the e-commerce platform implements personal recommendation of products based on browsing information of product. We propose a sentiment analysis method of product reviews based on the Joint Sentiment/Topic model, which can implement the personal recommendation of products based on the sentiment orientation of product reviews. Firstly, we build a sentiment dictionary for analyzing product reviews by integrating multiple external sentiment dictionaries. Secondly, we give a method to mark the sentiment polarity of the product reviews text. It can tag the sentiment polarity of the product reviews text to generate prior knowledge for the Joint Sentiment/Topic model. Finally, we give the formula for calculating the value of sentiment orientation on product reviews based on the Joint Sentiment/Topic model. Experiments show that the proposed method can effectively obtain the sentiment orientation of product reviews, making the product recommendation of the e-commerce platform more scientific and reasonable.