Microsoft Demos Real Time Language Translation Technology

by Joshua Gulick — Monday, November 12, 2012, 11:20 AM EDT

Although you may not agree after getting off the phone with your bank, speech recognition technology really has come a long way in the past 20 years. There’s still plenty of room for improvement, however, and Microsoft has been working on technology that is making for some big changes in real-time speech recognition. Microsoft’s Chief Research Officer, Rick Rashid, recently provided a pretty impressive demonstration of its capabilities.

The new speech recognition technique, developed by Microsoft with the University of Toronto, is called Deep Neural Networks. Based on human brain behavior, it, the technique provides significantly better accuracy than older techniques – in fact, it reduces the error rate by more than 30 percent, according to Rashid.

Rashid points to Microsoft’s research as moving us toward the universal translator from Star Trek before the 22nd century, which explains his (darn cool) necktie. In his demonstration, everything Rashid said was displayed as text while he spoke (with only a slight delay). The accuracy of the speech recognition is noteworthy, but so is Microsoft’s language translation technology, which turned his recognized text into Chinese and spoke audibly to the audience at the Microsoft Research Asia’s 21st Century Computing event in Tianjin, China.

Tags: Microsoft, speech recognition, Language Translation

Microsoft Demos Real Time Language Translation Technology

Login with Social Media or Manually