Google updates technology behind voice search to improve speed and accuracy
Google’s Voice Search services have garnered a lot of fame over the last couple of years, not just because it can do voice searches in a more accurate way than competitors, but because the service predicts a ton of things before you ask it anything through Google Now. Google’s main business is really to be a search engine, so accurate results are necessary for users to rely on its technology when in need. Today the company has announced new changes that should make its service even better.
Google has just updated the technology behind its voice search services with the idea of providing users with added speed and more accuracy. This new service relies on Connectionist Temporal Classification (CTC) and sequence discriminative training techniques. Yeah, we know that statement didn’t really help much in clarifying what exactly changed, but just to give you an idea, this would be pretty much the third change that Mountain View has done in its service over the last five or so years. While previous change from Gaussian Mixture Model (GMM) to Deep Neural Networks (DNN) allowed Google to better assess the sound a user was producing in order to provide better speech recognition, CTC is said to be much faster and more accurate. Yes, the more we write, the more confusing it is, so let’s just quote Google on a few more things:
“Our improved acoustic models rely on Recurrent Neural Networks (RNN). RNNs have feedback loops in their topology, allowing them to model temporal dependencies: when the user speaks /u/ in the previous example, their articulatory apparatus is coming from a /j/ sound and from an /m/ sound before. Try saying it out loud – “museum” – it flows very naturally in one breath, and RNNs can capture that. The type of RNN used here is a Long Short-Term Memory (LSTM) RNN which, through memory cells and a sophisticated gating mechanism, memorizes information better than other RNNs. Adopting such models already improved the quality of our recognizer significantly.”
The update should have already reached your iOS and Android devices by now, so let us know what you think in the comments.