Google today announced that it has made improvements to its voice search service on Android and iOS. With new improvements, the voice search tool is now faster and offers accurate results, even when used in noisy environments. Also Read - Google might be working on a 'Find My' network clone for Android users
The search giant explained what all it had done for improving the voice search functionality. It uses Connectionist Temporal Classification (CTC) along with sequence discriminative training techniques for better neural network acoustic. Google claims these new models are special extension of recurrent neural networks (RNNs) that offers improved accuracy, especially in environments with a lot of ambient noise. Also Read - PUBG New State receives over 17 million pre-registrations as closed alpha testing ends
With the new technology, voice search engine can quickly predict phonemes and also differentiate between letters that are similar sounding. For instance, the word “museum” in phonetic notation is broken down into /m j u z i @ m/. Also Read - Apple CEO Tim Cook claims iOS is more secure than Android
While the sounds made by “u” and “j” can be a little difficult to separate, the feedback loops helps the software to take all sounds and letters at once. To achieve this, Google is using the newly developed memory cells that work with a complex gating mechanism. Besides making voice dictation and search on Android better, it requires less computing power, thus also improving the battery life at the same time.