GSLDA: LDA-based group spamming detection in product reviews

Abstract

Online product reviews are becoming increasingly important due to their guidance function in people’s purchase decisions. As being highly subjective, online reviews are subject to opinion spamming, i.e., fraudsters write fake reviews or give unfair ratings to promote or demote target products. Although there have been much efforts in this field, the problem is still left open due to the difficulties in gathering ground-truth data. As more and more people are using Internet in everyday life, group review spamming, which involves a group of fraudsters writing hype-reviews (promote) or defaming-reviews (demote) for one or more target products, becomes the main form of review spamming. In this paper, we propose a LDA-based computing framework, namely GSLDA, for group spamming detection in product review data. As a completely unsupervised approach, GSLDA works in two phases. It first adapts LDA (Latent Dirichlet Allocation) to the product review context in order to bound the closely related group spammers into a small-sized reviewer cluster, and then it extracts high suspicious reviewer groups from each LDA-clusters. Experiments on three real-world datasets show that GSLDA can detect high quality spammer groups, outperforming many state-of-the-art baselines in terms of accuracy.