Despite the criticism, one of the features of Windows Phone 7 that impressed me is the text entry mode and its ability to correct the typing mistakes with surprising accuracy.
Here’s an early demo at MIX10 with Joe Belfiore:
“The technology originated as a brainstorm between members of our Mobile Communications Business and several Microsoft Research folks. They wanted to determine if the principles of machine learning could be applied to improving finger input on the soft keyboard. The idea being that by taking a lot of data that shows what corrections are typically made when entering text, you can teach the system to anticipate and auto correct. The more data you have, the more accurate the system becomes.
The team didn’t limit themselves to this and worked together to explore everything from typing intelligence and the underlying architecture, to UX and on-screen interaction. The MSR team put together a finger approach not unlike that used for speech recognition determining the right kind of data collection/evaluation and building models to help establish a data driven mindset. With immediate results from the first prototyping showing 30-40% reductions in input errors they knew they were on the right path.”
The result is what you see today in Windows Phone 7:
[via: Next at Microsoft]