Microsoft Speech Recognition is as Accurate as Humans

According to a research and paper published at Cornell University Library, a group of Microsoft Engineers have reported their system attaining  a word error rate (WER) of 5.9 percent. It is a number which is as good as achieving Human Parity in Conversational Speech Recognition.

To reach such level of accuracy, researchers at Microsoft deployed neutral networks to store sufficient data called as training sets. It helped the system recognize patterns from human input.

Want a Free Website
Microsoft researchers from the Speech & Dialog research group include, from back left, Wayne Xiong, Geoffrey Zweig, Xuedong Huang, Dong Yu, Frank Seide, Mike Seltzer, Jasha Droppo and Andreas Stolcke. (Photo by Dan DeLong)
For Microsoft Technology and Research: A research team photographed in Microsoft’s Building 99 in Redmond, Wash. on Thursday, October 13, 2016. Photo by Dan DeLong

“This accomplishment is the culmination of over twenty years of effort,” said Geoffrey Zweig, who manages the Speech & Dialog research group.

Extremely excited, the team now hopes to achieve even better results in the future, seek better-higher level of accuracy and ensuring that speech recognition works truly in real life situations, in all walks of life as technology evolves further.

“We’ve reached human parity,” said Xuedong Huang, the company’s chief speech scientist. “This is an historic achievement.”

Perhaps the future is going to be an era where computers will have to understand humans rather then other way around.

Via: TheNextWeb Microsoft

Want a Free Website