Restricted Boltzmann Machines for Modeling Businesses

摘要

User-generated business reviews have a significant impact in decision making of consumers. It would be useful to predict if a business is good for children or to predict its ratings using only text reviews. In this paper, an approach on modeling consumers reviews using a probabilistic model, Restricted Boltzmann Machines (RBMs), is proposed. We apply RBMs for non-linear feature extraction on Yelp data set consisting of raw data for businesses. Obtained results using RBMs modeling outperform results using traditional classifiers.