There's nothing as embarrassing as waking up to realize you drank too much the night before, and sent out lots of texts and Twitter tweets while you were doing that. You could spend a whole hazy day apologizing to people you wronged while you were buzzed.
As it turns out, the University of Rochester could soon prevent that kind of embarrassing weekend. They're using machine learning to help identify when someone is drinking and, at the same time, posting tweets.
Their algorithms learn from experience. The school's researchers first spent 12 months tracking geo-coded tweets, meaning they could tell where a person was tweeting from, like home or a bar.
Then they got a lot of help from people online. These online volunteers, referred to as Mechanical Turk, checked for words associated with drinking, like "beer" and "drunk." They read the tweets to confirm that the person sending the message was actually getting drunk, not just talking about it, and where they were at the time.
The University of Rochester has an academic reason for conducting this research. They're using their findings to improve their machine learning algorithm. And they hope to find out more about drinking behaviors and their impact on your health.
What would be really useful is if that machine learning one day is turned into a smartphone app. When it senses you're about to embarrass yourself, "beep."