Keyboard technology may not seem as exciting as faster microprocessors, massive displays, or ever-decreasing form factors, but in many ways it's just as important to computing. Seattle-based Swype is trying to leave its mark on the evolution of user input by making "pecking" at keys obsolete.
Instead of having to find and press on-screen keys one by one, Swype simply has users slide (or swipe) their fingers across the screen. Its algorithm does its best to figure out what you were trying to write, then fills it in for you. With a growing number of handsets shipping without a physical keyboard, this software could boost typing productivity and data usage by mobile phone users. Best of all, it doesn't have to replace the existing keyboard paradigm, meaning users can still peck if they like.
However, one big hurdle in the race to get Swype on every new handset is competition from all sides. Big companies like Apple, Palm, Google and RIM have invested in their own software-based keyboard solutions, while some competitors have working versions that accomplish what is effectively the same thing. Those companies also have their own patents and algorithms that help the software figure out what word you were really trying to type in. Swype's creators think they have found the sweet spot of having a product that's ready for mass market now, and that can evolve with its users over time.
Touch and go
Swype's technology was originally envisioned as a way to improve text input for disabled users. Those with limited dexterity are able to use Swype's system more easily than a traditional on-screen keyboard. It's also set up to support gesture tracking using Web cams, and with pointing devices like infrared remote controls, meaning it can be used on most hardware built within the last 10 years.
Swype's co-creator Cliff Kushler concocted it as an out-of-retirement project, and a follow-up to his previous co-invention T9--the text prediction algorithm that can be found in more than three billion mobile phones. Swype is trying to go beyond that though; following mobile phones the company has set its sights on tablet PCs, in-store kiosks, gaming devices and even televisions--basically, anything without a physical keyboard.
While Swype madeback in September of 2008, it won't be showing up in consumer devices until close to the end of this year. And of those devices, Swype is trying to hone its focus on mobile phones first.
For the past eight months the company been beta-testing Swype to a group of fewer than 100 users on Windows Mobile devices. It's also being readied for Symbian and Google's Android, the latter of which recently launched it's own on-screen keyboard solution in a system software update. It even runs on Microsoft's Surface. Swype CEO Mike McSherry and team, who I met with in their Seattle office last week, have been pushing at OEMs to be the default soft-keyboard in every future handset they ship. Swype's current iteration can be installed on-top of the existing keyboard that ships with Windows Mobile, but the big idea is to get it on the phone before the users get their hands on it.
While the design for Swype is still being fine-tuned, McSherry says that each carrier and OEM wants to re-skin it to match the brand. For example, T-Mobile could give the keyboard a pink hue to match its branding, and add special function keys that get to carrier-specific services. McSherry says that Swype's design allows for a greater customization without damaging the underlying user technique or reliability since the gesture zone for each letter will remain the same.
Swype's reference design has undergone multiple revisions, the latest of which adds a help guide that gives users tips on how to use it. This could prove useful as McSherry says Kushler has hidden all sorts of "Easter egg" gestures inside of Swype, including ones that let you draw symbols that get transcribed into complete words such as "circle," "infinity", and "arrow." Similar to Palm's Graffiti handwriting recognition software, this gives users shortcuts that can speed up the process of typing out complete words on top of Swype's letter-to-letter input method.
An evolving dictionary and skill set
What may be just as neat as Swype's process of text entry are the company's plans to build a rich database of user-generated words that periodically gets updated to all Swype users. This will let the software more easily guess misspelled words, or offer auto completion suggestions. It also cuts down on users having to teach their devices words with awkward spellings.
To do this, the company is adding top trending brands, movies and TV shows both from consumer indexes as well as Web sources like Twitter. However, the company is being careful about adding too many words, which can slow down the utility of its suggestion and autocorrect features. "Over-indexing on names is an issue," says Aaron Sheedy, Swype's senior vice president of business development. "Potentially it's a 65,000-word dictionary So maybe we're killing off 'pontiferous,' the 64,999th word to make room for something that will be used more. We want it to be a better use case scenario."
Besides Twitter, the company has also tapped some of its beta testers by using a logging system that sends user-created words and phrases back to Swype so they can be added in a global dictionary. This, however, isn't something that's slated to make it to the final product. "There are privacy concerns with that (system)," explains Sheedy. He says that while that may work for some common words, you get into a gray area with things like user names, passwords, and URLs.
Privacy concerns aside, during the beta test, this logging has shown Swype certain user metrics that give the team confidence that Swype has staying power. "It measures words per minute accuracy and tracks usability," says McSherry. "Quantitatively users are doing over 40 words per minute ... we can increase their WPM accuracy too. Within the first hour people are doing 30 WPM. Some people do 40, some even do 50."
And accuracy is a big deal. McSherry says that with a typing system like this even 90 percent doesn't cut it. "That means one out of every ten words you type is wrong, and you have to go back and correct it," he says. "We're shooting for 95 to 97 percent or more."
McSherry says he doesn't expect users to learn some of the more complex gestures that have been built into Swype, but that those who take the time to learn them can increase their speed even more, especially younger users. "You come up with T9 and text messaging and (kids) invented a new vernacular to evolve the technological opportunities ... It is an amazing way to watch a 13-year-old grab (Swype) and start flying with it off the bat."
Competition and IP control
While the future looks bright for Swype, critics have claimed the company's fundamental technology is quite similar to that of past competitors.
One of those is ShapeWriter (formerly Shark), a technology from IBM that made waves in late-2004 with many of the same promises. It too lets users write out words without picking up their finger (or stylus), and type in excess of 60 words per minute. Up until mid-2008 the use of the technology had remained fairly low-profile, but then in short succession it was released as a download for Google's Android, then on the iPhone and Windows Mobile. Like Swype, though, it still has not shipped as the default keyboard solution on a consumer handset.
There's also Dasur, a company that's got a solution for Windows Mobile devices that includes a slide-based typing feature, text prediction, and control over words in its dictionary. It's the only one of the bunch that's actually charging for the technology so far, with a $40 per device licensing fee.
In addition, Swype faces competition from Nuance Communications, which has its own T9 successor called xT9. This solution has both an on-screen keyboard that users can peck at, along with a handwriting recognition area where they can write in text and have it converted into characters or entire words. This may be different from Swype and ShapeWriter's solutions, but it's got the marketing force of T9 behind it--a predictive system that billions of users are familiar with.
The deciding factor in the race to be the top gesture-based onscreen keyboard could come down to a single, seven-letter word: patents. All three companies have various patents that refer to gesture-based on-screen typing. Swype holds five of its own, along with some dozen held by Kushler. The earliest of those was filed back in 2001. In comparison, ShapeWriter has three, including two that focus specifically on shorthand gesture-based shortcuts, and Dasur has one issued in mid-2007 that's nebulously described with "a user input mechanism is provided that allows a user to enter words as gestures on a virtual keyboard presented on the display device."
McSherry would not go into specifics on Swype's intellectual property strategy, but believes that between the patents--including the algorithm that includes prediction, and its differentiation from the others, that it can more than hold its own.
Swype will be rolling out a larger beta test later this year. In the meantime the company has a sign-up page to be notified when it's available for public download. To fill that in you'll need a normal keyboard though.