When the initial version of Google Voice Search was released by the speech recognition team, it was done so with a structure intended to “work fast” and “work right.” The input recognition had very broad settings that allowed the program to work immediately for almost any users, and covered different paces of speaking, accents, tone of voice, and more. Now, however, Google is taking their voice search a step further by adding personalized recognition.
As announced on the Google Mobile blog, the new version of Voice Search is available for download in the Android Market, and features the addition known as “personalized recognition.” The personalized recognition technology will allow Google to learn how you, specifically, pronounce words, further compensating for the exact way you speak. The program will learn more over time, and will develop more rapidly if you use it more often, but users should be able to see at least some improvements from the get-go.
The recognition software is also getting better at recognizing names. Most common names will be recognized immediately, although users will still notice that complex surnames provide an issue.
The improvement to the technology end doesn’t actually slow down the voice searches, either; quite to the contrary, it speeds them up. While the feature was already fairly lightweight, users on 3G networks should notice a speed increase.
User experience may also be enhanced by a new UI, which allows users to access a help page, cancel the search once initiated, and enter a settings page. From the settings page, the user can choose whether a content filter should be placed on search, whether offensive words should be blocked from voice recognition, and whether or not to enable personalized recognition.
The recognition data is saved to the user’s Google account, meaning that it has the potential to make voice features smarter across all associated devices.