Researchers at MIT have trained an algorithm that works on deep learning, to understand sarcasm through emojis. Dubbed DeepMoji, it used 1.2 billion tweets that contained some combination of the 64 most used emojis on the internet.
First it was taught how to predict which emoji were being used to express specific emotions and then it learnt how to identify sarcasm using examples from a vetted set of data. The researchers actually found DeepMoji can trump humans in spotting sarcasm and other similar emotions on Twitter.
The machine identified sarcasm correctly 82% of the time, while an average human could only muster 76% against the bot. Nonetheless, Gary King from the Institute for Quantitative Social Science at Harvard University is less convinced of this bot's value. He opines that if sarcasm is so nuanced that a human reader would miss it, it wouldn't matter in the longer run.
Either way, we have interesting times ahead with bots understanding the fine art of sarcasm!
