A survey on opinion spam detection methods

Since the past decade, fake Reviews also known as Opinion spam has plagued the e-commerce sector around the world. Opinion spam is considered extremely harmful as it can be used to control the sentiment of a product or service, which in turn can be used to damage the sales and reputation of a company. Throughout the years, extensive research has used Natural language processing for extracting textual features and use them with various machine learning algorithms for opinion spam detection. Majority of the reviewed literature has focused on supervised learning techniques using artificially crafted datasets. The purpose of this paper is twofold: to analyze the various machine learning techniques that have been proposed in the extant literature for detecting opinion spam and compare their accuracies, to provide further insights for future researchers in the field of opinion spam detection. This survey has concluded that semi-supervised techniques using multi-aspect features of reviews, reviewers, and products can provide a better result in spam detection. Furthermore, the lack of accurately labeled datasets presents a major challenge in the field of Fake review detection