Microsoft Wows with Machine Speech Translation in Real Time
December 11, 2012
Once again, we are catching up to our science fiction. The Next Web informs us about a recent leap in the field of machine translation with, “Amazing: Microsoft Turns Spoken English into Spoken Mandarin—in the Same Voice.” The article includes a nine-minute video slice of a presentation by Microsoft’s Rick Rashid that is well worth the viewing time (though the exciting part really starts about half-way through.) The video begins with a brief recap of the history of machine transcription and machine translation. Writer Alex Wilhelm tells us:
“In the video, the speaker explains and demonstrates improvements made to the machine understanding of his English words, which are automatically transcribed as he speaks. He then demonstrates having those words translated directly into Mandarin – if it’s actually Cantonese I’ll punish myself – text.
“This is when the fun begins. Microsoft, he says, has taken in oodles of data, and can thus have that translated Mandarin spoken. And the final kicker: he has fed the system an hour’s worth of his voice, and thus the software will speak in Mandarin, using his own tones.”
I would like to point out here that, despite the write-up’s title, “his own tones” does not quite equate to “the same voice.” It is close, though. Rashid attributes the leap to the development of Deep Neural Networks, a technique patterned after human brain behavior by researchers at Microsoft Research and the University of Toronto. The shift is indeed very impressive, and makes a future where we can all understand each other seem closer to possible.
We would be remiss, however, if we failed to mention that Google can still claim some advantage in this realm. Its Google Translate has been shown to generate more accurate text translations than Bing Translator. (See here and here for comparisons.)
So, Google, when do you debut your instant speech translation software? We can’t wait to see what you come up with.
Cynthia Murrell, December 11, 2012
Sponsored by ArnoldIT.com, developer of Augmentext