Twitter bot helps Chicago find dirty restaurants
Pedestrians walk past the Fogata Village restaurant in Chicago's Pilsen neighborhood.
For an increasing number of Chicagoans who get food poisoning, a very unpleasant experience has been followed with a pleasant surprise.
After getting sick from eating at a restaurant and complaining about it on Twitter, they get a message from the Chicago Department of Public Health. And, for an increasing number of those tweets, the department has followed up with investigations and fines.
Earlier in August, Chicago health officials published a report touting the success of their campaign, which employs a self-learning algorithm to mine Twitter for messages that may indicate that someone got sick from eating at a Chicago-area restaurant.
In 2013, the algorithm, and subsequent inspections, resulted in 21 restaurants being closed down and 33 others receiving citations for serious violations, Chicago officials say.
“We know a majority of food-borne illnesses never end up getting reported to public health, whether locally or nationally,” says Bechara Choucair, the Commissioner of the Chicago Department of Public Health.
Considering food-borne illnesses can cost $2 to $4 billion annually, and can affect 55 to 105 million people every year in the U.S., finding out about what’s causing them is an important public health issue.
“We knew that we needed to figure out a more effective way of listening to residents and really act on their complaints. And when we realized that the majority of people are not picking up the phone and letting us know that they’re getting a food-borne illness, but still going and talking to Twitter, we tried to figure out how can we leverage that,” Choucair says.
Choucair employed volunteer coders to create a Twitter bot that looks for messages geocoded to Chicago, which include the term “food poisoning.” Health officials then gave the algorithm feedback on the messages it was finding, and it learned to more correctly identify potential cases for investigation.
The algorithm alerts officials when it finds a potentially relevant tweet. A health worker then writes back to the tweeter and asks the to fill out an online complaint form to launch an investigation.
The program started in March of 2013, and by now, the number of investigations sparked from Twitter (while still about 10 percent of the total number of cases each year) is growing.
“What’s equally important is the feedback we’re getting from those residents. And they’re really excited to know that their local government is listening to them, but not only just listening, but also acting on their complaints and trying to prevent, and really making sure that they do have healthier lives in Chicago,” Choucair says.
Other health departments have also had similar ideas for mining online social networks for potentially troublesome restaurants.
In New York, health officials are keeping track of Yelp reviews, while Boston is trying its own experiment with mining Twitter.
Chicago health officials have been consulting with their counterparts in those two cities, so there may soon be Twitter bots for Boston and New York, as well.