The rising influence of user-generated online reviews has led to growing incentive for businesses to solicit and manufacture deceptive opinion spam—fictitious reviews that have been deliberately written to sound authentic and deceive the reader. Recently, Ott et al. (2011) have introduced an opinion spam dataset containing gold standard deceptive positive hotel reviews. However, the complementary problem of negative deceptive opinion spam, intended to slander competitive offerings, remains largely unstudied. Following an approach similar to Ott et al. (2011), in this work we create and study the first dataset of deceptive opinion spam with negative sentiment reviews. Based on this dataset, we find that standard n-gram text categorization techniques can detect negative deceptive opinion spam with performance far surpassing that of human judges. Finally, in conjunction with the aforementioned positive review dataset, we consider the possible interactions between sentiment and deception, and present initial results that encourage further exploration of this relationship.

Consumers' purchase decisions are increasingly influenced by user-generated online reviews. Accordingly, there has been growing concern about the potential for posting deceptive opinion spam—fictitious reviews that have been deliberately written to sound authentic, to deceive the reader. But while this practice has received considerable public attention and concern, relatively little is known about the actual prevalence, or rate, of deception in online review communities, and less still about the factors that influence it.

We propose a generative model of deception which, in conjunction with a deception classifier, we use to explore the prevalence of deception in six popular online review communities: Expedia, Hotels.com, Orbitz, Priceline, TripAdvisor, and Yelp. We additionally propose a theoretical model of online reviews based on economic signaling theory, in which consumer reviews diminish the inherent information asymmetry between consumers and producers, by acting as a signal to a product's true, unknown quality. We find that deceptive opinion spam is a growing problem overall, but with different growth rates across communities. These rates, we argue, are driven by the different signaling costs associated with deception for each review community, e.g., posting requirements. When measures are taken to increase signaling cost, e.g., filtering reviews written by first-time reviewers, deception prevalence is effectively reduced.

We investigate the efficacy of topic model based approaches to two multi-aspect sentiment analysis tasks: multi-aspect sentence labeling and multi-aspect rating prediction. For sentence labeling, we propose a weakly-supervised approach that utilizes only minimal prior knowledge—in the form of seed words—to enforce a direct correspondence between topics and aspects. This correspondence is used to label sentences with performance that approaches a fully supervised baseline. For multi-aspect rating prediction, we find that overall ratings can be used in conjunction with our sentence labelings to achieve reasonable performance compared to a fully supervised baseline. When gold-standard aspect-ratings are available, we find that topic model based features can be used to improve unsophisticated supervised baseline performance, in agreement with previous multi-aspect rating prediction work. This improvement is diminished, however, when topic model features are paired with a more competitive supervised baseline—a finding not acknowledged in previous work.

Consumers increasingly rate, review and research products online. Consequently, websites containing consumer reviews are becoming targets of opinion spam. While recent work has focused primarily on manually identifiable instances of opinion spam, in this work we study deceptive opinion spam—fictitious opinions that have been deliberately written to sound authentic. Integrating work from psychology and computational linguistics, we develop and compare three approaches to detecting deceptive opinion spam, and ultimately develop a classifier that is nearly 90% accurate on our gold-standard opinion spam dataset. Based on feature analysis of our learned models, we additionally make several theoretical contributions, including revealing a relationship between deceptive opinions and imaginative writing.

Press: ABC News, Businessweek, NPR, The New York Times.