Crowdsourced product and service reviews are a great way to gauge how good something is before you buy, but how do you know you aren't being had? The proliferation of for-pay opinion spamming in recent years has made it difficult to know if a five-star review is actually genuine. Luckily, spammer groups may have gotten a lot easier to spot, thanks to a new study by University of Illinois researchers and partially supported by a Google Faculty Research Award. The study, entitled Spotting Fake Reviewer Groups in Consumer Reviews, aims to uncover opinion spam using a new relation-based algorithm called GSRank (Group Spam Rank).
The approach uses a number of features of a group and its members — including how closely together the members posted their reviews, similarity of review content, how frequently the group has worked together — and observed relationships between group spam and products, member spam and products, and individual spam and group spam, in order to make inferences about the group. The result is an algorithm that "consistently outperforms all existing methods" for detecting bogus reviews. While moderators probably aren't too keen to solve the study's complicated eigenvalue problem every time they want to check for astroturfing, hopefully we'll see the algorithm packaged into something more user-friendly in the future.