Predicting helpful product reviews

Millions of users are today posting user-generated content online, expressing their opinions on all manner of goods and services, topics and social affairs. While undoubtedly useful,user-generated content presents consumers with significant challenges in terms of information overload and quality considerations. In this paper, we address these issues in the context of product reviews and present a brief survey of our work to date on predicting review helpfulness. In particular, the performance of a variety of different machine learning approaches is evaluated on four large-scale review datasets drawn from the TripAdvisor and Amazon domains. Our findings highlight some interesting properties of this task from a machine learning perspective and demonstrate that author reputation, the sentiment expressed in reviews and review length are among the most effective predictors of review helpfulness.