How to interpret the helpfulness of online product reviews: bridging the needs between customers and designers

Abstract

Helpful reviews are the valuable voice of the customer which benefit both consumers and product designers. On e-commerce websites, consumers are usually encouraged to rate whether a review is helpful or not. As consumers are not obligated to vote reviews, usually only a small proportion of product reviews eventually receive a voting. Also, existing evaluation methods that only use the review voting ratio from customers as the helpfulness are often not consistent with the designers' rating on reviews in interpreting customer needs and preferences. Thus, in this paper, the focus is on how to automatically build the connection between online customer's voting and designer's rating and predict the customer reviews' helpfulness based on the review content. We start the study by building a mapping to express product designers' rating using online helpfulness voting. Further, we propose to utilize regression algorithm to predict the online review's helpfulness with the help of several categories of features extracted from review content. Our experimental study, using a large amount of review data crawled from Amazon and real ratings from product designers confirms the effectiveness of our proposal and shows some very promising results.