| Latest in Translation |
|
11 November 2005, for Science Grad student Stan Jou was mouthing Mandarin Chinese, but no sounds issued from his mouth. Instead, a robotic voice from a speaker spoke for him, using inputs from electrodes glued to his cheeks and throat. The words, in English or Spanish, were part of a press conference last week at which computer scientist Alex Waibel of Carnegie Mellon University in Pittsburgh, Pennsylvania, and others showed off their latest toys for speech recognition and translation. The electrodes on Jou's face picked up movements of his face and throat muscles. Software turned them into words, which were then translated. So far, the system can only recognize about 15 phrases. But Waibel predicts that someday people will be able to have face-to-face conversations in alien tongues without the sounds of their original words getting in the way. The researchers are also developing goggles displaying simultaneous translations of a talk. And they've built directed speakers that can pinpoint a person in a crowd and deliver a translation as if it were being whispered in the ear. Waibel's software for translating spoken language is some of the best in the world, says Satoshi Nakamura of the Advanced Telecommunications Research Institute in Japan, but he doubts such a program will make it to the marketplace in this decade. Some of this technology could require more-or-less permanent attachments to the listener. But, says Waibel, "I think someday people will accept having a few electrodes implanted in their cheek." |