Moderating Effects of Time-Related Factors in Predicting the Helpfulness of Online Reviews: a Deep Learning Approach

Abstract

Given the importance of online reviews, as shown by extensive research, we address the problem of predicting the helpfulness of online product reviews by developing a comprehensive research model guided by the theoretical foundations of signaling and social influence theories. We use review order and time interval to incorporate the moderating effects of the time-related variable on the reviewer’s valuation of products and the related details they provide. Applying deep learning techniques in text processing and model building on a dataset of 239297 reviews, the empirical findings represent strong support of the proposed approach and show its superior performance in predicting review helpfulness compared to current approaches. This research contributes to theory by analyzing online reviews from the points of two well-known information processing theories and contributes to practice by developing a model to sort the newly posted reviews.