If you find Foursquare and Yelp restaurant reviews too subjective, here's a somewhat more robust method for harnessing social data to inform your dining choices. By collating tweets about unhappy meals and their locations, a group of University of Rochester researchers have been able to generate a map of potentially risky places to eat. More specifically, the nEmesis system looks for tweets that suggest their author is "likely suffering from a foodborne illness" and then tie the tweet's location to the address of the restaurant visited.

The ingenuity of the present system is in the algorithms and human-guided machine learning that are employed to filter out the noise and pinpoint relevant tweets.

After working through some 3.8 million tweets from more than 94,000 New Yorkers, the Rochester team found a valid correlation between their system's predicted health score and official data from the US Department of Health and Mental Hygiene. That four-month trial has shown nEmesis has the potential to be an inexpensive complement to traditional food safety measures. In fact, due to its ability to rapidly identify potential hotspots for food poisoning, it could be used to guide inspectors in conducting their surveys.