URI_Research_Magazine_Momentum_Spring_2015_Melissa-McCarthy

Equal Rating Opportunity

by Chris Barrett

Americans annually spend some $260 billion online. And, 90 percent of shoppers consult product reviews before handing over their credit card. To ensure the honesty of those reviews, University of Rhode Island (URI) computer engineering Associate Professor Yan “Lindsay” Sun and her students developed a simple online tool rooted in signal processing theory. The tool allows shoppers to input the website address of a page containing product reviews on Amazon. In a few seconds, the site returns a creditability rating and details how it reached that score presented on a scale of 0 to 1, with 1 extremely reliable.

on Amazon and it’s not a transparent process how they check their reviews,” Sun says. “With this customers can make their own judgments.” The tool works by sending the reviews through an algorithm that identifies patterns. After studying 240,000 reviews, Sun and her students determined reviews should not follow a pattern. There is no reason all reviews posted on Saturday should be positive. Nor should all positive reviews be from newly registered users, or come from a tiny subset of geographic locations. The team coined its findings the Equal Rating Opportunity principle in a play on words with the Equal Employment Opportunity principle in human resources. The beauty of the principle lies in its simplicity. Traditional checks of reviews involve combing massive amounts of data across thousands of webpages. The process seeks to find anomalies like users that always post positive reviews or always for

one company. But that method relies on gaining access to non-public information and could take hours to process. The Equal Rating Opportunity principle instead needs only information any normal web surfer can see. The principle caught the attention of the Rhode Island Hospitality Association, which contacted the professor after reading about her work in URI’s QuadAngles in 2014. Executives asked if the same principle could be applied to finding fraudulent reviews on Yelp that unfairly cost businesses like hotels and restaurants sales. Intrigued, Sun tasked a graduate student to investigate and, found, in short, yes. By tweaking the algorithm, Sun and her student are building a system where business owners could analyze a negative review and get a scientific result about its validity. Armed with that information, based on URI’s independent algorithm, the business could petition Yelp to remove the review.

For example, in a blatant case, a

“Without this tool you have to rely

Page 52 | The University of Rhode Island { momentum: Research & Innovation }