Sign In

Communications of the ACM

ACM TechNews

A Major Mystery Behind Microsoft's 'brain-Like' Speech-to-Speech Translator

View as: Print Mobile App Share:
Screenshot from a Skype text chat translator product.

Microsoft CEO Satya Nadella this week introduced speech-to-speech translation technology as Skype Translate.

Credit: DYC Software

New speech-to-speech translation technology from Microsoft becomes better at learning a language as it learns additional languages. For example, when the technology learns English and is then taught Mandarin, it becomes better at English, and it will become even better at both languages when it is taught Spanish.

No one knows why the technology is brain-like in the sense of its capacity to learn, notes Microsoft CEO Satya Nadella, who introduced Skype Translate at the recent Code Conference. Skype Translate is powered by a process that is a mix of speech recognition, machine translation, and speech synthesis. "It's not just about daisy-chaining these three technologies and bringing it together," Nadella says. "In fact, it's this deep neural net that you build that synthesizes a model to be able to do speech recognition."

He says Microsoft will have more details closer to the beta test about how many languages Skype Translate will support and how it becomes aware of what languages it is dealing with in each conversation.

From Network World
View Full Article


Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA


No entries found