It's notoriously tough to translate sarcasm into the written word, but with the help of technology and science, it soon might get a little bit easier to tell when someone's being sarcastic with their emoji.
According to MIT Technology Review, researchers that initially looked to create an algorithm to filter out racism accidentally stumbled upon a way to identify sarcasm.
Using deep learning, which looks for repeated patterns in huge amounts of data, the researchers noticed that emoji were an indicator of strong emotion. Looking closer at emoji specifically, the team pivoted to focus on the tiny icons and the messages they accompanied.
AdvertisementADVERTISEMENT
"Because we can't use intonation in our voice or body language to contextualise what we are saying, emoji are the way we do it online," Iyad Rahwan, associate professor the MIT Media, wrote. "The neural network learned the connection between a certain kind of language and an emoji."
The algorithm looked at the 64 most commonly used emoji and scanned through 55 billion tweets to train the system to connect the emoji with various sentiments. After running through all the data, Rahwan and his team found that the algorithm could match various emoji with messages, even if sarcasm was at play.
Of course, MIT wanted to compare its robotic emoji maestro to humans. During a test, researchers found that humans could correctly identify sarcastic tweets 76% of the time. DeepMoji, which is what the team decided to call the new system, identified 82% of the tweets. DeepMoji can also suggest emoji to go along with any tweet, so if you're ever at a loss when choosing the perfect one, turning to this super-smart algorithm could help you make that very important decision.
It's good news for the impending robot takeover. According to Rahwan, tech like this could allow computers and other AI to sense human emotion, be it through a sarcastic compliment or frustration at a complex computer program, and toss out an emoji to either comfort their human compatriots or really annoy them.
"If machines are going to cooperate with us, then they're going to have to understand us, and emotion is really hard," Rahwan said.