Android App SwiftKey publicly launched today to compete with Swype and other assisted typing tools by essentially creating predictive text on crack.
It has done this by creating a machine learning engine that goes much further than the (highly annoying IMHO) predictive text on the iPhone and other smartphones. Applying supercomputing power to analyze relationships between words, it predicts what your next word will be, even before you start typing it.
It can also bounce between different languages with ease and learn a user’s writing style over time. The latter involves monitoring the way you combine words and reviewing past content that’s stored on your phone (e-mails, etc.).
The company claims that a third of the words you type can be predicted before you type the first letter, and a majority predicted within a few letters. This improves typing efficiency by 50 percent it says.
This isn’t the sexiest area of mobile development, but it’s important. Beyond being a useful app, this is something that will be a link to better product development and adoption of the mobile Web. As we’ve mentioned, one of the things holding back mobile Web adoption — in favor of the warm and fuzzy world of apps — is user friendliness.
A lot of this comes down to the discoverability of content, which stems from two main factors. The first is the lack of content that is optimized for mobile, creating a wild west environment reminiscent of the desktop Web of 15 years ago. The second is the nonintuitive, but requisite, finger tapping of URLs and search terms.
The first reason will be alleviated over time by HTML5 and cost effectiveness of mobile Web development (spawning more and better content). But the second gets back to the main point of this post: the development of more intuitive search inputs. This comes with the premise that search — like it is on the desktop — will be the front door to the mobile web.
This will include voice and visual search, as evidenced by the investments being made by vested interests (read: Google). But typing — or more accurately, finger tapping — will continue to exist and will need to get more user friendly to push mainstream adoption of search (both Web and apps).
SwiftKey is a step in that direction. And if the engine behind its predictive text is as good as it seems, it will transcend keyboards and make its way to inputs like voice and speech-to-text processing. Watch closely; this will be a big area of mobile development, if a non-sexy one.