Google Assistant Personal Speech Recognition helps you learn the keywords you use most often
Google Assistant – the company’s smart, voice-controlled digital helper – is soon getting the ability to better understand your frequently used words and names according to a recent report.
With its ‘Personalise speech recognition’, Google Assistant will be able to store and process this data purely on your smartphone – not sending to the cloud – which could then improve its ability to respond to and understand what you’re saying.
This news comes through 9to5Google in a report which details some of the changes it has spotted deep within the code of the Google app on the Play Store.
Having decomplied the app, the site reports that this ‘Personalised speech recognition’ will store recordings on your phone/device, and that this data can be deleted at any time, simply by switching off the feature.
This is an enhancement of the ability Google Assistant already has to recognise your voice when you use the ‘Hey Google’ wake phrase, expanding it to words that you use, and actual commands and requests that you make.
As reported by 9to5Google, the recordings of these requests will then be analyzed on the device, which in turn helps transcription and makes responses more accurate and relevant in future.
The report also states that it’s likely this will be an opt-in feature when it lands officially, it won’t just be turned on by default. So if you don’t feel comfortable with recordings of your requests being kept anywhere, it shouldn’t just do it by default. You’ll have to switch it on for it to work.
If and when this new capability arrives, it likely won’t do so with a huge reveal, but likely a Google blogpost at time of release. We’ll keep an eye out and keep you updated when and if that happens.
keyword spotting speech recognition github
google voice recognition
small-footprint keyword spotting using deep neural networks
speech recognition in ai
speech recognition example
google keyword spotting
keyword spotting papers with code
speech recognition machine learning