For some frequent texters, a message isn’t complete without a few carefully chosen emojis. The colorful pictographs can be used to soften blunt language or add humor to dry sentences. Tech companies like Apple have already developed ways to “emojify” the words we type, but Dango may be the first translation tool to treat emojis like a complex language.

According to FiveThirtyEight, Dango’s programmers used deep learning algorithms to teach the app how people communicate with emojis in real life. Like the upcoming emojification tool for iOS, Dango started out offering only simple word-by-word translations like a dog emoji for "dog" or a smiley face for "happy." But when it came to interpreting the meaning of a string of words into relevant emojis, the tool fell flat.

To improve Dango’s emoji literacy, the team developed a recurrent neural network containing more than 300 million emojis from 180 million messages from platforms like Twitter and Instagram. Neural networks are designed to work like structures in the brain, and they can be used to teach translation tools not just the meaning of a word but how it relates to others (Google recently started using this technology to improve Google Translate).

Dango’s intuitive nature saves users time crafting texts, in addition to introducing meaningful emojis they may not have thought to add on their own. For example, for messages that might be construed as sarcastic it suggests either a "thumbs up" symbol or a side-eyed "unamused" face. Texts that contain the word "haters" generate the sassy "nail polish" emoji and the word "Beyonce" produces a crown with a bumblebee, a.k.a. "Queen B."


Dango

For every emoji Dango uses effectively, there are some that are likely to get lost in translation. This is further complicated by the fact that humans don’t always agree on what the same emoji symbolizes, even when looking at the same design from the same platform. But thanks to deep learning, the more people communicate with emojis, the more adept Dango will be at using them. The app is now available for Android with an iOS version on the way.
.
[h/t FiveThirtyEight]

Know of something you think we should cover? Email us at tips@mentalfloss.com.