Windows Phone voice recognition receives improvements to be faster and more accurateAlex Wagner - Editorial Director of News and Content
Remember that video demonstration of improved voice recognition in Windows Phone 8 that leaked out earlier this year? It looks like Windows Phone users will finally be able to try out those improvements for themselves, as Microsoft announced today that it's made voice search on Windows Phone twice as fast and 15 percent more accurate.
Over on the official Bing blog, Microsoft has revealed that its Bing team has been working with the Microsoft Research group on Deep Neural Networks (DNNs) technology, which can detect patterns in a way similar to biological systems can. Microsoft was able to its Bing index to help the DNNs learn more quickly than usual, which lead to Bing voice search being able to recognize speech in a manner more closely to the way that humans do.
In addition to its work with DNNs, Microsoft says that it's made tweaks to Bing voice capabilities that will help it to cut its response time in half and better ignore background noise. The company claims that these tweaks have helped to improve Bing's word error rate by 15 percent.
Voice search is becoming an increasingly large focus of mobile operating systems and the apps that run on them. Apple recently announced that it's made a number of enhancements to its Siri voice assistant in iOS 7, and we've seen Google improve its voice controls in Google Now and its other apps. Now its time for Windows Phone users to enjoy an improved voice search experience as well, which is great news considering the voice enhancements that we've seen popping up on other platforms as of late.
Microsoft says that its been rolling out the updated voice capabilities to Windows Phone users in recent weeks, so anyone rocking a Windows Phone should already have access to the speedier voice capabilities while composing a text or performing a search. For a more detailed dive into DNNs and a demo of the speedier voice controls, you can check out a video demo at the link below.