Published in News

Microsoft acquires SwiftKey

by on03 February 2016

Predictive input market competition at all-time high

Software giant Microsoft has recently acquired mobile artificial intelligence startup SwiftKey, a company that has developed a reputation for its predictive key input method software for iOS and Android mobile devices.

With today's acquisition, competition in the predictive input race is heating up to an all-time high on all three major corners of the mobile device landscape. Back in early 2014, Google aquired DeepMind, another London-based AI company, for about $400 million. In October 2015, Apple acquired Perceptio, a smartphone image recognition startup. At the time, Perceptio was developing technology that would allow smartphone makers to build advanced AI systems "without needing to share as much user data" in the cloud.

SwiftKey is another London-based company, founded in 2008 by Jon Reynolds and Dr. Ben Medlock, and currently employs a staff of over 160 people. In 2014, the company pulled in $17.5 million in funding, switched its consumer keyboard app from a $3.99 paid download to free. The cost switchup allows SwiftKey to reach a much wider user base, and came around the same time Apple opened up iOS 8 to allow the installation of third-party system wide keyboards.

The company's algorithm used in SwiftKey are primarily rooted in the machine-learning subfield of artificial intelligence. It adapts to a user's personal use of language so it can better predict their next likely inputted words. The software's "Prediction Engine" currently learns from SMS conversations, Facebook status updates, Gmail inboxes, Twitter updates and RSS feeds and currently supports 106 languages.

The latest version of the software, SwiftKey 6, was released in November 2015 and now includes double-word prediction, to predict a user's next two words at once. Other improvements include a complete overhaul of the settings menu in the style of Google's Material Design and a redeisgn of the emoji panel.

swiftkey 6 two word prediction

SwiftKey 6 using Google's Material Design interface

In October 2015, SwiftKey announced a new project called SwiftKey Neural Alpha, the world's first predictive input algorithm using an artificial neural network (ANN) and inspired by the structure and operation of the human brain. Through machine learning based on enormous amounts of language data, the company's neural model can meaningfully capture the relationship between words. This is a big upgrade from its current SwiftKey "classic" app, as the neural language model allows predicting words that have never been seen in the context of a given sentence during the learning phase.

The software is available as an early-stage project on SwiftKey Greenhouse and can be downloaded for Android 4.4 KitKat and above.


swiftkey classic vs neural

SwiftKey Classic (left) "n-gram" technology vs SwiftKey Neural (right) artificial neural network

Now that Microsoft has acquird SwiftKey, there really is no saying in what direction the new user interface will go. The software's adoption of Google's Material Design is just two months old, and we can't help but question any executive decisions from Microsoft to keep one of their biggest competitors' latest groundbreaking UI improvements on top of such an important mobile software acquisition. Whether SwiftKey becomes redesigned to incorporate a more Windows Mobile-themed redesign is currently in question, but we would expect another change coming soon.

As Fortune Magazine notes, the continuous interplay between SwiftKey's predictive algorithm and what a user is typing gives a more convenient experience to those looking to increase messaging efficiency at a rate much quicker than native word-only predictive models on iOS and Android can provide.

Microsoft's acquisition SwiftKey will be announced Wednesday, with early estimates at a $250 million purchase price. But sources close to the deal have said this price is a tad high and will probably be finalized a bit lower, with more information following the announcement.

Last modified on 03 February 2016
Rate this item
(4 votes)

Read more about: