Algorithm reads tweets to figure out which restaurants make you sick
By scanning social media, the app can find likely hot spots.
By Katherine Noyes
Food poisoning can strike anywhere hygiene standards are lax, but researchers have developed a new app that uses machine learning to help minimize the number of people affected.
One out of every six U.S. residents gets food poisoning each year, and when they do, many of them write about it on Twitter. That’s where nEmesis comes in. Developed by computer-science researchers from the University of Rochester, the software uses natural language processing and artificial intelligence to identify food poisoning-related tweets, connect them to restaurants using geotagging and identify likely hot spots.
The researchers developed their app by analyzing almost 4 million tweets generated by people in the New York City metropolitan area in late 2012 and early 2013. Included in the software are deep-learning algorithms trained to recognize key phrases — “I feel nauseous,” for instance.
More recently, the researchers tested the app in Las Vegas through a collaboration with that city’s health department. Specifically, Las Vegas began incorporating nEmesis results in its targeting of restaurants to inspect on any given day.
For three months, the system automatically scanned an average of 16,000 tweets from 3,600 users each day. The researchers then used those tweets to generate a list of the highest-priority restaurants for inspections.
“Each morning we gave the city a list of places where we knew that something was wrong so they could do an inspection of those restaurants,” said Adam Sadilek, a researcher who worked on the project at the University of Rochester and is now at Google Research.
As a result, there were 9,000 fewer food-poisoning incidents and 557 fewer hospitalizations in Las Vegas during the course of the study, the researchers estimated.
The team presented the research at the 30th Association for the Advancement of Artificial Intelligence conference in February. It received an Innovative Applications of Artificial Intelligence award from the association.
The same approach could also be used in a variety of other ways, Sadilek said.
“This happens to be restaurants, but the method can also be used for bedbugs,” he explained. “Similarly, you can look what people tweet about after they visit their doctor or hospital. We’re just beginning to scratch the surface of what’s possible.”