• Reuters, Wednesday, March 28, 2007 12:30 PM

Google envisions a world of statistical machine translation, whereby computers instantly translate everything from documents to the things people say to each other in chat rooms across the Web. As ever, the search king believes automated programs are the way to go, not (in this instance) human linguists.

The company doesn´t even plan to use linguists to program its grammatical rules and dictionaries. Instead, it will feed documents already translated into two languages into its program, which will learn rules and patterns for future translations, based on the data it´s been fed.

Franz Och, a German who heads Google's translation effort at its Mountain View, Calif. HQ, says to create the program, he feeds hundreds of millions of words from parallel texts, such as Arabic and English, into the computer, using United Nations and European Union documents as key sources. "The more data we feed into the system, the better it gets," he said. The focus thus far is on the work, not the ad model. "As part of a general Google philosophy, once it's really useful and it has impact, then there will be found ways how to make money out of it," Och says.

Read the whole story at Reuters »