Google has improved its voice search, which is more accurate now, and the credit for it goes to the advances in artificial intelligence, the Internet firm said Thursday. Now, the voice search will be more accurate even in noisy surroundings, and execute the required analysis in quick time.
Google voice search more accurate
In a blog post, Google Speech Team members Ha?im Sak, Andrew Senior, Kanishka Rao, Françoise Beaufays and Johan Schalkwyk wrote that they have been successful in developing even better neural network acoustic models by making use of Connectionist Temporal Classification (CTC) and sequence discriminative training techniques. “These models are a special extension of recurrent neural networks (RNNs) that are more accurate, especially in noisy environments, and they are blazingly fast!”
Google Speech Team members wrote that the new models not only require much less computational resources, but they are also more accurate, robust to noise and faster to respond to voice search queries. The team members noted that the new models are working on the Google’s iOS and Android apps, and also on dictation on Android that works inside some third-party apps.
For differentiating between similar sounding letters, the acoustic models make use of feedback loops. With feedback loops, the software takes all letters and sounds at once just like it takes one breath to say the word. The newly developed memory cells work with a sophisticated gating system to make this possible.
Growing importance of Google Voice Search
Google reported improvements in voice search twice this year so far, indicating the investments made by it in the underlying technology. RNNs are an increasingly popular approach to doing Deep learning, which is a type of artificial intelligence. And, it is widely believed that Google has a ‘deep bench in deep learning,’ says VentureBeat.
In coming years, the importance of speech as an input to search the web will increase by large. Andrew Ng of Baidu, known for his work on the so-called Google Brain, expect speech and images to be used for 50% of the searches and queries within five years.
Realizing this, many tech firms are working on improving their voice recognition capabilities including Facebook, Apple and Microsoft. A few months back, Facebook acquired Wit.ai, a speech recognition company, to up its skills in the area as well.