In fiction Douglas Adams had his babel fish and Star Trek had the communicator, but Microsoft Research has been demoing an actual real-time English to Mandarin translation engine that works within seconds.
Microsoft's chief research officer Rick Rashid showed off the technology at the company's Asian 21st Century Computing conference last month, and has released a video showing the engine in operation. Within three seconds of him speaking in English, the system converted his voice into Mandarin and displayed a written Chinese translation.
The breakthrough that made this possible lies in a technique called deep neural networks, developed by Microsoft Research and the University of Toronto. The system takes English words, seeks a Chinese translation of them, and then matches those words to a recorded set of Chinese spoken words.
Current translation systems – even Microsoft's, Rashid explained – have error rates of around 25 per cent. The new system cuts this by over a third, taking errors from one word in four to one every seven or eight, which while not perfect, should avoid a hovercraft and eels situation.
"There's much work to be done in this area, but this technology is very promising and we hope in a few years we'll be able to break down the language barriers between people," he told the conference, while the systems translated. "Personally I believe this will lead to a better world."
There's certainly a crying need for translation systems between English speakers and residents of the Middle Kingdom – this looks to be the Chinese century and there's a real shortage of translators. Mandarin, Cantonese, and their many dialects are very difficult languages to learn for non-native speakers, since they require a tonal finesse seldom used in English and other languages. This system could be a godsend.
Judging from the applause, rather than polite giggles, from the Chinese speakers in the audience, the system works as advertised. If any Mandarin speakers notice a howler, or two please let us know. ®