4 ms·
There is a technological aspect, and that's putting the voice processing back into the device itself. We can still get the benefit of mass machine learning on
by nitrogen 11y ago
There is a technological aspect, and that's putting the voice processing back into the device itself. We can still get the benefit of mass machine learning on failed voice samples by asking users to submit the failed sample only, playing it back for them so they know what is being sent, and deleting the sample once it has been processed. Then an updated speech model can be sent to the device. Or, just do the retraining on the device, too.