Microsoft has shown off the Windows Phone 8 keyboard, giving a comprehensive look at what it offers over Windows Phone 7, as well as the prediction technology it utilizes. In particular, it takes some time discussing how the keyboard was made for "real people," better accommodating language usage and slang, for example. It then goes on to discuss how it deals with the "fat finger problem," complete with a video demonstration.
First up is Word Flow, which is an improved version of Quick Correct with a new, more encompassing name. The feature is designed to reduce the number of typos a user experiences while better accommodating the way people naturally talk, including slang. Microsoft selected 600k of the most commonly used words out of 2.5 billion that were reviewed, giving Windows Phone 8 an autocorrect accuracy percentage of 94-percent.
It has achieved this by loading its integrated dictionary with frequency data that supplies information on how often each word is used. This helps Windows Phone 8 predict what word the user is going for, reducing the frequency of wrong words. One of the examples Microsoft presents is how Windows Phone predicts what word to choose when a user types the letters "h-a-p-p". There are many words that can be selected, such as happen and happy. Because "happy" is more commonly used than "happen," it presents that first.
In addition, the mobile operating system accommodates slang and real-world usage, such as nonstandard spellings. It does this via a dictionary unlike the Webster sitting on your bookshelf. The dictionary of nonstandard words and slang is derived from anonymously collected typing data from Windows Phone users. Microsoft utilized social networks and Wikipedia, as well, for common and new slang terms, which were automatically gathered and manually reviewed.
And what of the fat finger problem? Microsoft dealt with this issue, which results as often from small displays as it does large fingers, by implementing hit targets. A hit target is a sensitive area around a key that changes based on Word Flow's prediction of what you're trying to type. This helps it predict what key you were aiming for rather than which one you actually hit. You can see an example of this technology in the video above.