You may not have noticed it but the Google application on your Android smartphone or iPhone silently got updated to serve you better. The speech recognition is somehow better than it used to be and unlike before, it pretty much gets what you asked on your first try and processes it faster.
Google has replaced the underlying core technology used for voice search from the Deep Neural Networks (DNNs) it’s been using since 2012 to the more accurate and fast recurrent neural networks (RNNs). As a result, searching for specific commands and just about anything you fancy on the Google application is faster and more accurate and you won’t need to walk out of a noisy room just to ask Google what the scores of your favourite football team are. The Google app hears you even in noisy environments. Google says that it even uses lower computational resources.
Interested in diving deep into how the teams at Google managed to do this? Hit the source link below.
Source: Google Search blog