Scientists create twitter app to fight food poisoning
New York's University of Rochester scientists have created an algorithm that analyses tweets to determine which restaurants are prone to serving up foodborne illnesses, so users can steer clear.
The system, called nEmesis, detects tweets sent from restaurant locations and then tracks user tweets for the next 72 hours, looking for words describing classic food-poisoning symptoms. If any users tweet about being ill, nEmesis captures this data and flags the restaurant.
According to The Atlantic, nEmesis mined 3.8 million tweets from 94 000 unique users in New York City over a four-month period, tracking 23 000 restaurant visitors and spotting 480 reports of possible food poisoning. Plus the results correlated with health department scores - the more tweets associated with being sick, the more likely a restaurant had been given a failing grade.
Tweeting your gastrointestinal woes? Maybe not for you, but the new service is part of a new scientific trend that puts all that oversharing to good use. In fact, cocreater Adam Sadilek, who now works at Google, hopes to see nEmesis in the form of a mobile app, much like Germ Tracker, his previous project that uses similar technology to alert users to flu outbreaks in their social media circles.
When developing Germ Tracker, Sadilek and a team of researchers used Twitter to track how factors such as social status, exposure to pollution and even taking the bus or going to the gym influence one's health. By following thousands of Twitter users in New York, many of whom sent tweets that were location tagged, over a period of a month, researchers were able to estimate interactions between users and their environment.
A new study published this summer also used Twitter as a tool to help scientists understand not only what we eat but why. University of Arizona researchers, after analysing tweets from the study's 50 Twitter-happy volunteers, found that food choices were typically driven by cost and convenience rather than hunger. The findings appear in the Journal of Medical Internet Research.