In truly monumental research, Nabil Hossain at the University of Rochester took Twitter and machine learning, and smashed them together to determine drinking habits across specific geographic locations. Hossain and his team rounded up thousands of geotagged posts that appeared on Twitter between July of 2013 and July of 2014 across the state of New York, and then parsed them down to only examine tweets that contained alcohol-related keywords (think “beer keg,” “drunk,” and the like).
This ultimately gave them a database of about 11,000 tweets to work with, which the researchers passed through Amazon’s Mechanical Turk crowdsourcing service. Every single one of those thousands of tweets were examined by three human “Turkers,” who answered three questions about the posts:
- Does the tweet make any reference to drinking alcoholic beverages?
- If so, is the tweet about the tweeter him or herself drinking alcoholic beverages?
- If so, is it likely that the tweet was sent at the time and place the tweeter was drinking alcoholic beverages?
The answers derived from these questions gave the team a sense of whether or not the tweets came from folks who were drinking or drunk. Then, going even further with their research, Hossain and his team looked into where the tweeters were when they were imbibing (or at the very least, drunk tweeting). By the end of their efforts, they’d managed to create an algorithm that theoretically can pinpoint whether a user was at home when drunk tweeting with up to 80-percent accuracy.
By combining these two determinations, Hossain says that his research group can now determine where New Yorkers prefer to drink, and by extension, identify their favorite watering holes. “Our results demonstrate that tweets can provide powerful and fine-grained cues of activities going on in cities,” Hossain’s team notes, and tracking this sort of activity could help keep drinkers out of harm’s way.
So don’t be embarrassed by your drunk tweets (not for this reason at least). After all, they could save your life.
