They say that sarcasm is the lowest form of wit, but that hasn’t stopped a research team in the Netherlands training AI to recognize it. Matt Coler, a researcher at the University of Groningen’s speech technology lab, told The Guardian: “We are able to recognize sarcasm in a reliable way, and we’re eager to grow that. We want to see how far we can push it.”
He continued: “When you start studying sarcasm, you become hyper-aware of the extent to which we use it as part of our normal mode of communication. But we have to speak to our devices in a very literal way, as if we’re talking to a robot, because we are. It doesn’t have to be this way.”
In developing their AI, it was reaffirmed for the researchers exactly how much non-verbal social cues like delivery and facial expressions matter when it comes to detecting sarcasm. They found that numerous cues were important in helping the AI distinguish between those being sarcastic and those speaking sincerely.
Introducing Mustard: The Sarcasm AI
Researchers across the US and Singapore built up a database (known as Mustard) full of video clips showing instances of sarcasm from TV shows like Friends and The Big Bang Theory.
Xiyuan Gao, a PhD student, then described how researchers utilized Mustard to train the neural network on how to detect sarcasm. After training the AI on these scenes, which was fed through the machine in both audio and textual formats, it was found that the AI could detect sarcasm 75% of the time.
Meanwhile, Gao added that the accuracy of the machine could be increased by implementing more visual cues into the AI’s training like smirks and eyebrow raises, “Are we going to have a machine that is 100% accurate?” Gao asked. “That’s not something even humans can achieve.”
Either way, at least this is less terrifying than AI deliberately rejecting instructions.
Featured Image: CBS