Dango is the latest app release from Whirlscape, creator of Minuum, that suggests meaningful, and apt, emojis for messages exchanged within a messaging app.
The app works on any messaging platform in smartphone devices and appears as a pink cake icon on the messaging interface. This Dango icon can then be tapped to make it suggest appropriate emojis based on the user’s input.
“Dango appears in any messaging app, and helps you quickly find emoji, stickers, and GIFs, by understanding what you type,” the Dango website describes.
Dango uses artificial intelligence (AI) to understand the context behind every message users type on their smartphone. This, in turn, helps the app suggest emojis that perfectly encapsulate the entire situation and not just a single word in the conversation.
Even more so, Dango is adapted to the current trends in social media and can suggest particular memes that not even its user has heard of yet.
This level of intelligence was acquired through the apps’ constant training in an artificial environment of deep neural networks.
Dango’s own network was primarily given a set of real-world emoji usage examples that were pretty much random to the AI. These examples would gradually increase by lifting real-world content from the internet to further increase the network’s comprehension and accuracy.
Through multiple deep learning simulations over a “top-of-the-line GPU,” the app readjusted its parameters accordingly to also find similar connections between emojis and words. As we all know, each individual emoji can be understood in a variety of ways, depending on its context.
This level of understanding creates a vast network of objective and subjective meanings affixed to each emoticon and thus, Dango’s intelligible framework is born.
Dango’s neural network will not stop “learning” on its launch, though, and will continue to update its repository according to the latest trends. Users, on the other hand, do not have to worry about slow loading times or privacy invasion as the app saves and accesses its contents locally, only updating as prompted but never transmitting a user's own personal messaging data.
One problem that arises from the app’s learning technique through real-world datasets is its capacity to acquire certain trends that are generally negative, ie. racist or sexist, as was seen in Microsoft’s recent Twitter AI, Tay.
In a report, Will Walmsley, Dango co-founder, explains that the company is doing its best to prevent such occurrences from happening.
“Bad associations are a real, and subtle, concern since AI tools like Dango reflect back our human best and our worst. We’ve in the past blacklisted certain associations and we’ll keep doing this … but it will be an ongoing challenge to get right,” the co-founder adds.
Dango is publicly available for download by Android users on the Google Play Store while iOS devices will be getting the app sometime soon. Current app ratings at the time of this writing are 4.2 out of 5 stars, based on 265 reviews.
Dango's app trailer can be viewed below:
