Google said it has built a better neural network that is making its voice search work faster and better in noisy environments.
"We are happy to announce that our new acoustic models are now used for voice searches and commands in the Google app (on Android and iOS), and for dictation on Android devices," Google's Speech Team wrote in a blog post on Thursday. "In addition to requiring much lower computational resources, the new models are more accurate, robust to noise, and faster to respond to voice search queries."
In 2013, Google brought the same voice recognition tools that had been working in Google Now to Google Search.
Along with being able to find information on the Internet, Google Voice Search also was able to find information for users in their Gmail, Google Calendar and Google+ accounts.
At the 2013 Google I/O developers conference, Amit Singhai, today a senior vice president and Google Fellow, said the future of search is in voice. For Google, he said, future searches will be more like conversations with your computer or device, which also will be able to give you information before you even ask for it.
The company went on to make it clear that it would continue to focus on voice search.
And this week's announcement backs that up.
Google explained in its blog post that it has updated the neural network it's using for voice search. A neural network is a computer system based on the way the human brain and nervous system work. It generally uses many processors operating in parallel.
The improved neural network is able to consume the incoming audio in larger chunks than conventional models without performing as many calculations.
"With this, we drastically reduced computations and made the recognizer much faster," the team wrote. "We also added artificial noise and reverberation to the training data, making the recognizer more robust to ambient noise."
Patrick Moorhead, an analyst with Moor Insights & Strategy, said if Google wants more people using voice search they need to continue improving its accuracy.
"Very few people rely on any smartphone voice products," he said. "They only will rely on them when the accuracy gets to 99%. Imagine if your keyboard only was accurate 95% of the time, meaning every five keys out of 100 were inaccurate. So this is a nice improvement, but until the complete user experience gets to 99% it won't really make a difference in adoption."