Scientists at University College Dublin (UCD) have developed a system intended to improve the helpfulness of online customer reviews, and have been awarded a distinguished paper award at the British Computer Society's annual Artificial Intelligence Conference for their efforts.
Barry Smyth, Michael O'Mahony, Maurice Coyle, and Peter Briggs of UCD, as well as Pierre-Antoine Champin, a visiting professor from the University of Lyon, examined 225,000 reviews by 45,000 distinct reviewers on 70,000 hotels on the travel website TripAdvisor. According to their research, as much as 65% of reviews did not provide useful information. In response, they applied machine learning techniques to discover the factors that make for a helpful review.
Their technique analyzes reviews that have generated feedback and uses a model generated by those reviews to determine the worth of reviews that have generated little or no feedback.