In recent news, a computer algorithm can now tell how drunk you based on the tweet you just posted. A team of scientists from the University of Rochester had devised a computer program that can analyze your tweet and tell you how drunk you were when you wrote it.
Who said that science can’t be fun? Well, a group of researchers from the University of Rochester proved once and for all that you can have tons of fun and laughs even inside the lab. Recently, the dastardly team managed to devise a nifty computer algorithm that can tell how intoxicated you were last time you wrote on Twitter.
Why did they do that? Because it’s fun and it’s in the name science. Leaving the joke aside, for some time now, computer researchers wanted to find out how many people use social media platforms after they get plastered. Furthermore, the team also wanted to see how social media channels like Twitter, Facebook or Snapchat can influence our drinking habits.
How was this done? Well, it wasn’t easy, that’s for sure. First of all, the team of scientists had to look for the so-called drunk posts. But where to look? For their scientific endeavor, the team chose Monroe County and New York City.
After selecting the most favorable locations, the team started to look through tons of tweet to identify those who have been written while under the influence of alcoholic beverages.
They didn’t have to look for too long. The whole searching process was narrowed down to tracking down a few keywords like “beer”, “party”, or “alcohol”.
The next step of the process involved some pretty heavy lifting, so the boys asked the help of Amazon’ Mechanical Turk. With some extra muscle, the team of scientists managed to discover no less than 13.000 drunk tweets.
And, according to their statements, the computer algorithm, using geotags and keywords, managed to identify whether a tweet was posted from someone’s home or the immediate vicinity of his or hers home in 80 percent of cases.
Moreover, based on their preliminary results, the team of scientists also said that more people are drinking at home and posting on Twitter in New York City than in Monroe County.
In the end, in order not to enrage anyone, the team said that the results do not reflect the fact that New Yorkers might have an alcohol-related issue in comparison to those from Monroe County. The whole purpose of this experiment was to develop a new deep-learning algorithm, one that can push the boundaries of computer sciences even further.