Fact or factitious? Contextualized opinion spam detection

Abstract

In recent years, it has been shown that falsification of online reviews can have a substantial, quantifiable effect on the success of the subject. This creates a large enticement for sellers to participate in review deception to boost their own success, or hinder the competition. Most current efforts to detect review deception are based on supervised classifiers trained on syntactic and lexical patterns. However, recent neural approaches to classification have been shown to match or outperform state-of-the-art methods. In this paper, we perform an analytic comparison of these methods, and introduce our own results. By fine-tuning Google's recently published transformer-based architecture, BERT, on the fake review detection task, we demonstrate near state-of-the-art performance, achieving over 90% accuracy on a widely used deception detection dataset.