If you actively use voice search Google, you probably have noticed that in recent years the service began to give more precise answers in much less time.
Google has officially announced that it is connected with the introduction of new technology, which significantly improves the performance of voice search. Moreover, the system is now better versed in noisy places and hassle recognize your voice.
Also, the transition to the new technology will significantly reduce the computing resources consumed.
“Today we are pleased to announce that we have built an even more high-quality neural network acoustic models using the technology of the time and consistent classification connectionism discriminatory teaching methods. These models are a special extension for recurrent neural networks (RNNS), which are more accurate and incredibly fast “,
reported in the official blog of Google.
New acoustic model processing voice calls are already being used in the respective applications for Android and iOS.