Towards Amazon Fake Reviewers Detection

Online marketplaces such as Amazon allow people to share their experiences about purchased products using textual comments known as product reviews. These reviews have become a common tool that users rely on to get insights on the quality and functionality of products and services from online consumers. However, like any other online information, reviewers raise serious questions concerning the credibility and reliability, since anyone can post reviews, which might impact the reliability of the information. This paper tackles the phenomenon of Bulk reviewers. We first analyze a large dataset of reviews from Amazon aiming to spot bulk reviewers according to their behavior. We then apply a what-if analysis to assess the effect of bulk reviews on the online marketplaces using a metric called Net Promoter Score to measure the willingness of users to recommend products. Our Results reveal that bulk users (i.e., users that review multiple times) have same distribution of ratings as non-bulk users indicating that a bulk reviewer is not automatically a fake reviewer. Yet, we discover that bulk users do inflate NPS metric and thus contribute to overestimate the level of customer satisfaction.