New MIT Algorithm Uses Emoji to Learn About Sarcasm

Today Translations is confident that demand for emoji translation is set to grow
AFP

A new algorithm developed by MIT researchers is probably more emotionally intelligent than you are — on Twitter.

The algorithm, “DeepMoji,” used well over a billion examples of online communication to learn how emoji are used as a substitute for the tone and body language that brief online text messages cannot otherwise express. In doing so, it attained a level of accuracy that outstrips most human’s ability to do the same.

Originally, DeepMoji’s creators were simply trying to find a way to accurately identify racist messages on Twitter. In studying the subject further, however, it became clear that the artificial intelligence algorithm could not effectively do its job without a broader understanding of human social quirks. Foremost among them? Sarcasm.

In testing, AI trained with emoji were far better at identifying the intent behind a tweet than those that were not. In fact, it was consistently better than its own creators. In a test conducted using volunteers crowdsourced using Mechanical Turk, human subjects averaged a strong 76 percent success rate in identifying sarcasm. DeepMoji scored 82 percent.

The practical uses for such an artificial intelligence are many and varied. While it may prove an effective measure against the “hate speech” for which it was originally intended, it also carries the potential to help companies to more accurately identify attitudes toward their products. Furthermore, the ability to detect the general tone of public response to a certain brand could point toward major market shifts that would be of special interest to those marketing them — or their competition.

On a more elemental level, an artificial intelligence that can interpret human interactions through tone and context cues will naturally become better at communicating with humans in general. As A.I. continues to push the boundaries of neural network complexity, that will be as important for us, as it is for them — lest we end up in a :scream: :gun: :robot: scenario.

Follow Nate Church @Get2Church on Twitter for the latest news in gaming and technology, and snarky opinions on both.

COMMENTS

Please let us know if you're having issues with commenting.