Opinion Spam Detection in Online Reviews

Abstract

Online reviews are the most valuable sources of information about customer opinions and are considered the pillars on which the reputation of an organisation is built. From a customer’s perspective, review information is key to making a proper decision regarding an online purchase. Reviews are generally considered an unbiased opinion of an individual’s personal experience with a product, but the underlying truth about these reviews tells a different story. Spammers exploit these review platforms illegally because of incentives involved in writing fake reviews, thereby trying to gain an advantage over competitors resulting in an explosive growth of opinion spamming. The present study analyses and categorises the available literature on opinion spamming according to three detection targets: (1) opinion spam, (2) opinion spammers, and (3) collusive opinion spammer groups. The study further highlights and divides opinion spamming into three types based on textual and linguistic, behavioural, and relational features. Moreover, several state-of-the-art machine-learning techniques for opinion spam detection have also been discussed in the study. It concludes with a summary of the research articles on opinion spam detection and some interesting results to assist researchers for further exploration of the domain.

Machine Cleaning of Online Opinion Spam: Developing a Machine-Learning Algorithm for Detecting Deceptive Comments

“…However, attention has not been evenly distributed. Perhaps due to foreseeable financial consequences, for years, most scholarly and industrial efforts have been directed at detecting fake consumer reviews of products or services (e.g., Crawford et al, 2015; Jindal & Liu, 2007; Li, Chen, Liu, Wei, & Shao, 2014; Ott et al, 2011; Rastogi & Mehrotra, 2017). Discerning fraudulent positive or negative reviews from genuine consumer feedback becomes an issue at stake for both the selling and buying parties as well as opinion spam researchers.…”

“…Carefully crafted opinion spam is likely to show no obvious indications of deception in the text (Crawford et al, 2015). Human judgment is not reliable in this regard (Jindal & Liu, 2008; Ott et al, 2011), and even experts find it hard to detect fake reviews (Rastogi & Mehrotra, 2017). At the same time, the volume of messages (here, reviews) is vast (Crawford et al, 2015).…”

“…Since opinion spam detection is essentially a binary classification task— deceptive or truthful , automated supervised machine learning, in particular, has emerged as one of the most powerful methods in this field. However, undeniably, difficulty in obtaining a labeled (i.e., fake or genuine) ground truth dataset to train a machine-learning model for prediction makes “supervised” approaches less feasible for this task (Rastogi & Mehrotra, 2017). Opinions are “ubiquitous” online (Ott et al, 2011, p. 311) but are also contaminated (Jindal & Liu, 2007).…”

“…Inspired by the need for a reliable deceptive comment detection method, this study aims to develop an automated machine-learning technique capable of determining opinion trustworthiness in a comment section. In the process, we have created the first large-scale ground truth dataset consisting of 866 truthful and 869 deceptive comments on social issues (see Rastogi & Mehrotra, 2017, for discussion of the lack of ground truth datasets as one of the biggest challenges in this research area). The resulting algorithm is expected to provide a base methodology for further investigation.…”

Deceptive Opinion Detection Using Machine Learning Techniques

“…Spam reviews can be of different types. According to literature [4], opinion spam can be categorized into two types: positive opinion spam and negative opinion spam. Generally, reviews containing only advertisements or random texts are not usually considered as disruptive opinion spam and are much easier to detect by manual inspection.…”