Google’s machine translations aren’t perfect… but getting there
March 11th, 2010
Google has refined its translation tool to a point that “can make the language barrier go away,” as one of the principal scientists of the company’s machine translation team said. Now handling 52 languages, Google is yet again a visionary in an area most internet/computer companies have ignored over the years.
Remember those funny Babelfish translations you’d get at the dawn of the internet age, the computer translator that would give you “They are a small potentiometer, short circuits and a beer of malzes of the tea” for “I’m a little tea pot, short and stout”? Google has made those roundabout interpretations all but extinct.
How does machine translation work? And what makes Google’s so good?
Creating a translation machine has long been seen as one of the toughest challenges in artificial intelligence. For decades, computer scientists tried using a rules-based approach — teaching the computer the linguistic rules of two languages and giving it the necessary dictionaries.
But in the mid-1990s, researchers began favoring a so-called statistical approach. They found that if they fed the computer thousands or millions of passages and their human-generated translations, it could learn to make accurate guesses about how to translate new texts.
It turns out that this technique, which requires huge amounts of data and lots of computing horsepower, is right up Google’s alley.
Let’s be clear, no computer translation program will ever be able to capture the linguistic and cultural nuances beyond the text. Only a thinking human can interpret text that way, and as we’d always prefer, a professional one with lots of experience. Google recognizes this too, but for anyone needing a quick translation of a news article, Google translations certainly will capture the “essence” of the story.
The New York Times reports: click here.





