Precision technology Google's speech-recognition increased by 30% since 2012
2012 precision technology Google's speech-recognition increased by 30%. About it said a leading software engineer Jeff Dean at the conference Frontiers of AI, which takes place in California.
So a significant progress in technology has been achieved by adding a neural networks – systems that Google uses as part of machine learning. The mass application of neural networks to speech recognition was developed by a company in 2012.
In 2015 the head of Google, Sundar Pichai said that the error rate in speech recognition is 8%. For comparison, in September 2016 Microsoft managed to reduce this indicator to a record of 6.3%.