Microsoft announced that through research into Deep Neural Networks (DNN) it has doubled the speed, and improved the accuracy, of voice search on Bing when using Windows Phone. DNNs use a form of pattern recognition that better mimics the way the human brain works.
Anyone who has used any type of voice recognition software know that there have always been two issues that were at the center of the problem. Speed and accuracy. One without the other results in a mediocre experience at best.
Enter the researchers at Microsoft. The researchers at Microsoft Research in Asia and also the Conversational Systems Research Center in Redmond, WA have been working together to get vast improvements in both speed and accuracy of Bing Voice Search.
Using DNNs, a computational framework for automatic pattern recognition, the teams have been able to double the speed at which results are returned with a word error rate improvement of 15 percent. By refining the mathematical algorithms and using greater computational power and larger data sets, they have been able to reduce errors due to noisy environments and natural changes in voice (like yelling or changes from exertion).
The article at the source link is quite interesting, so hop over there to learn more about what this amazing technology can offer.