Can a flurry of negative Yelp reviews be the first sign that a restaurant is on deathwatch? The Washington Business Journal reports researchers at the University of Maryland's Smith School of Business have created software that claims to predict "with 70 percent accuracy" if a restaurant will close, based solely on its Yelp reviews. According to the report, researchers looked at nearly 2,500 restaurants — including 450 that had closed — in the Washington, DC area and combed through reviews to "identify linguistic patterns that foretold closure." The keywords were separated into categories, with positive markers like "food," "good," and "friend" among the words that suggested "Quality_Overall."
Restaurants high in "Quality_Overall" words "tended to survive at unusually high rates," according to researchers, with the text-based analysis more accurate than just a simple analysis based on star ratings. According to associate professor Anandasivam Gopal, a co-developer of the software, the program could predict whether a restaurant would close within the next three months seven out of 10 times.
Researchers have often turned to the user-generated Yelp for more than just restaurant recommendations: In recent months, health-care researchers have used the site to help track outbreaks of foodborne illness. According to the WBJ, the Smith School of Business "hopes to expand the research to look at the impact of Groupons or other discounts on a restaurant's business or possibly to examine health code inspection reports for restaurants." The Smith School of Business, it should be noted, currently has no reviews on Yelp.