At Microsoft’s 14th annual Computing in the 21st Century Conference in October in Tianjin, China, Microsoft’s Chief Research Officer Rick Rashid demonstrated the latest success of real-time translation from English into Mandarin Chinese. The video of the presentation is below; its transcript is now available online (“Microsoft Research shows a promising new breakthrough in speech translation technology” on Microsoft Next).
The success of the translation depended in large part on new developments in speech recognition software. According to Rashid, human speech is the single most important user interface and the most difficult for computers to process. Even the best speech recognition software systems to date have had error rates of some 20 to 25 percent.
Now, however, researchers at the University of Toronto and Microsoft Research have developed a new technique modeled on human brain behavior called Deep Neural Networks. This development reduces speech recognition errors by more than 30 percent, to rates like 1 in 7 or 8.
"While still far from perfect, this is the most dramatic change in accuracy since the introduction of hidden Markov modeling in 1979, and as we add more data to the training we believe that we will get even better results," said Rashid.
In the presentation, Rashid’s live speech was transformed into on-screen text, machine translated, and then “spoken” in Chinese to the audience by Rashid’s digital voice.
"It happens in two steps," explains Rashid. "The first takes my words and finds the Chinese equivalents, and while non-trivial, this is the easy part. The second reorders the words to be appropriate for Chinese, an important step for correct translation between languages."
We agree with Rashid, this technology is very promising and no less important for what it means for machine translation’s effect on global business.
Native Chinese speakers — in no less than 7 major dialects, of which Mandarin, Wu, and Cantonese are the most prominant — number approximately 1.6 billion worldwide. And although we’ve reported just recently on the Chinese government’s initiative to bring Chinese into schoolrooms worldwide, there’s still a need for tools that make multilingual communication and cooperation immediately possible to the business community, and not just in its printed form.
That’s true for whatever language is spoken, so let’s look forward to what Deep Neural Networks brings to the world.