Microsoft’s real-time speech translation breakthrough
MANILA, Philippines - On Thursday, November 8, Microsoft Research released a YouTube video and an explanation on the research blog regarding some breakthrough developments in speech translation.
In this case, from English to Chinese.
The 9-minute video shows Rick Rashid, chief research officer at Microsoft Research, during his October 25th presentation in China. He was explaining their machine translation improvements where his speech was transcribed and translated by the machine almost instantaneously, though not perfectly.
In the blog discussing the developments, Rashid explained, “We have been able to reduce the word error rate for speech by over 30% compared to previous methods.
"This means that rather than having one word in 4 or 5 incorrect, now the error rate is one word in 7 or 8.”
Perhaps the most intriguing portion of Rashid's presentation was a demonstration of the machine speaking a Chinese translation of his words using — as The Next Web put it — Rashid's own tones.
Rashid explains that Microsoft researchers built a text-to-speech system using a native Chinese speaker’s voice and combined it with Rashid’s voice properties, as based on an hour of his recorded English speeches.
“Though it was a limited test, the effect was dramatic,” Rashid writes, “and the audience came alive in response. When I spoke in English, the system automatically combined all the underlying technologies to deliver a robust speech to speech experience — my voice speaking Chinese.”
Rashid mentions that the system’s results aren’t perfect yet, but the promising nature of these developments may make it possible for the universal translators of science fiction to become reality, perhaps sooner than we think. - Rappler.com