Google's Voice Search Is Getting Way, Way Better

6,463

Filed to: VOICE RECOGNITION9/25/15 4:46am

Google's Voice Search Is Getting Way, Way Better

Google released its first neural network-powered voice recognition system in 2011 and it’s been slowly improving ever since. But now, it’s announced that the addition of recurrent neural networks will make far faster — and way more accurate.

In a blog post, the Google Speech Team explains that it’s added what are known as Connectionist Temporal Classification and sequence discriminative training techniques to its algorithms. If that doesn’t make much sense to you, here’s a straightforward explanation of how it works:

In a traditional speech recognizer, the waveform spoken by a user is split into small consecutive slices or “frames” of 10 milliseconds of audio. Each frame is analyzed for its frequency content, and the resulting feature vector is passed through an acoustic model... The recognizer then reconciles all this information to determine the sentence the user is speaking. If the user speaks the word “museum” for example - /m j u z i @ m/ in phonetic notation - it may be hard to tell where the /j/ sound ends and where the /u/ starts, but in truth the recognizer doesn’t care where exactly that transition happens: All it cares about is that these sounds were spoken.

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.

By introducing that ability to include information about sounds on either side of each snippet, the algorithms stands a far better chance of understanding what you say. In fact, Google claims that it makes voice search far more accurate, particularly in noisy environments, as well as helping to make it “blazingly fast.”

You don’t even need to do anything to take advantage of the improvement: The new neural network approach is already being used by the Google search app for iOS and Android.

[Google via Engadget]

6Reply

All replies

The following replies are approved. To see additional replies that are pending approval, click Show Pending. Warning: These may contain graphic material.

Show pending

mlk007Jamie Condliffe
9/25/15 5:11am

No surprise there. As someone who owns iPhone 6, Google now picks up my Indian accent and my kids British accent much better than Siri.

4

Reply
- Sir Vivormlk007
  9/25/15 5:48am
  
  Cortana only understands me when I’m drunk. Or what do I know, never tried it when sober.
  
  8
  
  Reply
- mrd1Sir Vivor
  9/25/15 6:24am
  
  Thank you! I’m going to borrow that as my excuse. I'm drinking because, otherwise, my phone won't work.
  
  5
  
  Reply
- tekkblade128mlk007
  9/25/15 7:50am
  
  For a frame of reference, are the accents you speak of thick or more subtle? Either way that’s pretty impressive for a program to be so adaptable.
  
  Reply
SkipErnstJamie Condliffe
9/25/15 7:42am

After watching the Hound demo, I have to guess they are using something like this too. The response times are very fast.

Reply
V@noJamie Condliffe
9/25/15 7:54am

Well, I don’t know...I just asked google “what’s up?” (as I like to ask Cortana), it showed words “whats up” at the bottom of the screen as if it understood me, however a second later it showed me search result for “axapta”.

Reply