Toronto-based Whirlscape develops deep learning emoji app

There’s a specific kind of frustration felt only by frequent emoji users.

It comes when the emoji seeker has to scroll for what feels like miles to find the tiny icon they desire. That’s the issue that Toronto-based Whirlscape has taken upon itself to solve with its new app Dango.

Dango is a floating assistant that works in all major messaging apps, delivering a selection of emojis from the 1,624 standardized options that are appropriate to the message’s sentiment and subject.

Though the idea may sound trivial or even frivolous, the tech behind it is serious stuff. Dango is powered by a neural network, a computational structure inspired loosely by the brain, with millions of adjustable parameters connecting in a manner somewhat akin to neurons.

When it was first created, those parameters were randomly initialized and Dango was fed millions of real-world examples of emoji use to learn the nuanced art of choosing emojis that match the message’s content.

Dango not only suggests single emojis, the app also provides GIFs, stickers and strings of emojis on a theme or subject. For example, if you write “GTFO,” it’ll provide two pointing hands a door. The program is also continuously learning, to make sure it keeps current on pop culture references.

To put Dango to work, all users have to do is tap on the pink square to the right side of its keyboard and select one of the options it provides.

Dango’s creator, Whirlscape, previously produced Minuum, an intelligent keyboard that aims to take up as little screen space as possible, and is backed by Y Combinator and BDC Capital. That project was first launched in 2013 on Indiegogo, doubling its goal of $10,000 in one day.

For now, Dango is only available for Android, but Whirlscape promises that the iOS version is coming soon.

This article was originally published on MobileSyrup