Description
The past two decades have seen the internet expand exponentially and the online shopping community has increased as well. Since there is an increased interest in purchasing items online, numerous methods were created to help consumers decide whether to trust the site they visit. One of these methods is reviews of a product of interest or a service being sold. Online review sites such as Yelp have been created where the community goes and rates products and service providers on their customer satisfaction. These reviews are then read by potential shoppers to help in their purchasing decisions. Many businesses now realize that keeping a good rapport on these online sites can lead to an influx of business which provides an edge over competitors. This new phenomenon in the world of competition has created an opportunity for fraudulent behavior. Individuals and institutions have begun leaving fake reviews with the intent to influence a consumer's decision. These writers are allegedly hired to create false reviews in order to help companies increase their rapport with consumers. The online community has been subjected to this behavior at an alarming rate. This thesis explores the ability of an automated tool to classify reviews as legitimate or illegitimate. The three main components in the analysis of reviews are the reviewer, the service or product being reviewed, and the review itself. One could attempt to detect false reviews by investigating some or all three components. The tool designed here tracks the character distributions of each review. These distributions are then compared to one another in an attempt to determine the validity of a review. Other information is also collected such as the amount of stars each reviewer awards to the store or service. This allows researchers to also investigate trends in how reviewers grant stars. The success of these approaches will be discussed along with other innovative methods already created to address this prevalent problem.
