A couple of fake reviews probably won't kill (or unfairly promote) somebody's product/hotel/restaurant - but sustained postings of fake reviews by large groups of colluding opinion spammers might succeed. To the rescue come software engineers from the University of Illinois in Chicago and Google - who this week reveal a software algorithm that can pinpoint review threads in which groups of fraudsters are trying to wrest control of online sentiment.
In a paper to be delivered at this week's World Wide Web 2012 conference in Lyon, France,
Illinois researchers Arjun Mukherjee and Bing Liu, alongside Google's Natalie Glance, found the task easier than they expected thanks to the spammers' mob-handed behaviour. They hired eight online review experts from ecommerce sites Ebay and Rediff and got them to assess the 'spamicity' of 2400 English language reviews as being 'spam', 'borderline spam' or 'non spam'.