Automatic speech translator
Status: Could appear in laptops or personal digital assistants by mid-2004
Some social pundits claim that communication via computer has hampered personal connections. But researchers at IBM are on the verge of using computers to bring people closer together with a system that translates spoken language on the fly.
The speech-to-speech effort started a couple of years ago “as an adventurous research project,” says David Nahamoo, manager of the human-language technologies group at IBM’s T. J. Watson Research Center in Yorktown Heights, NY. The group has now built a working prototype: a laptop computer uses speech recognition software to process spoken words into text; sophisticated translation algorithms convert the text into a second language; and then the computer uses text-to-speech technology to “speak” the translated phrase.
So far, the prototype works only for English and Mandarin Chinese. The IBM team chose these languages in part because so many people speak them; another reason is that they represent extremes of differences in just about any linguistic parameter-for instance, in “prosody,” or the meaning given to a word or sentence by the speaker’s inflection. The system works with phrases likely to be used in specific situations, namely, travelers ordering in a restaurant, navigating a city, or seeking emergency medical care. Project coordinator Yuqing Gao says the laptop prototype works well enough for two people speaking different languages to carry on a rudimentary conversation. A version of the system that runs on a personal digital assistant is already in development, and though the group isn’t planning work with other languages, the technology is language independent; once a need is identified, it can rapidly be applied to any pair of languages.