Google personalises Voice search for better accuracy

15 Dec 2010

Google has revamped its Voice search capability for Android devices by creating the ability for the app to learn the user’s voice and quirky speech patterns to conduct a more accurate search.

With Google Voice Search all users have to do is say what they’re searching for and a full search kicks in.

However, until now, the speech recognition software just identified the words and wouldn’t recognise one voice from another.

“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,” said Amir Mane, Google Voice product manager.

Personalised speech model

“If you opt into personalised recognition, we begin to associate the recordings of the words that you ask us to recognise with your Google account.

“We then automatically use these words to build a speech model specifically for you. This speech model enables us to deliver greater recognition accuracy. Although subtle accuracy improvements build over time,” Mane said.

The personalised recognition functionality is currently only available in English in the US but more countries will be added in the near future.

It will only work on devices with Android 2.2 or higher. The new app also includes improvements to name recognition and speed, especially over 3G and EDGE connections.

John Kennedy is a journalist who served as editor of Silicon Republic for 17 years