Google Uses Your Speech to Personalize Voice Search

Google said it is offering to link queries users speak into their Android smartphones to their Google Accounts to improve speech recognition for those users over time.

It's called personalized recognition, and it represents a graduation of Voice Search, which has become a popular search app on smartphones based on Google's Android operating system, as well as Apple's iPhone.

Google launched Voice Search more than two years ago with broad speech models that didn't account for users' age, accents and other characteristics.

"But we always knew we could build a more accurate model by listening to your voice, and learning how you-as a unique individual-speak. So today we're launching personalized recognition," Google Product Manager Amir Man??« and Glen Shires, a member of the technical staff at Google, wrote on the Google Mobile blog.

English speakers using Android 2.2 smartphones or later who opt into personalized recognition will have recordings of words they speak into their phones for Voice Search associated with their Google Account.

Google will use these words to build a speech model for the user to boost recognition accuracy. Ideally, Google Voice Search associated with users' Google accounts will improve the speech recognition capabilities when they conduct voice searches on their handsets.

Users may also choose to disassociate their voice recordings from their Google Account through the speech section in the Google Dashboard.

Support for other countries and languages is planned for the near future. Personalized recognition is not an over-the-air upgrade for Android 2.2 handsets.

Users must download the latest version of the Voice Search app from Android Market to begin personalizing speech recognition. eWEEK did so for a test version of the Samsung Nexus S and found the process to be quick and painless.

Google has shown an interest in boosting its speech and voice recognition capabilities of late. More than a week ago, the company acquired speech synthesis specialist Phonetic Arts, whose software samples human speech and tailors it for computers.

A person familiar with Google's plans said the Phonetic Arts capabilities did not come into play for personalized recognition.

"The Phonetic Arts team will join Google's existing efforts in speech technology to help drive innovation in the area of text-to-speech synthesis," a spokesperson told eWEEK Dec. 5.