Business: Do you wanna sell more? Discovering Topics, Sentiments and Prediction of Ratings

Abstract

In the era of Social Computing, the role of customer reviews and ratings can be instrumental in predicting the success and sustainability of businesses as customers and even competitors use them to judge the quality of a business. Yelp is one of the most popular websites for users to write such reviews. This rating can be subjective and biased toward user’s personality. Business preferences of a user can be decrypted based on his/ her past reviews. In this paper, we deal with (i) uncovering latent topics in Yelp data based on positive and negative reviews using topic modeling to learn which topics are the most frequent among customer reviews, (ii) sentiment analysis of users’ reviews to learn how these topics associate to a positive or negative rating which will help businesses improve their offers and services, and (iii) predicting unbiased ratings from user-generated review text alone, using Linear Regression model. We also perform data analysis to get some deeper insights into customer reviews. © 2017 Association for Computing Machinery.