31 May, 2005

Google Translator: The Universal Language:
At the recent web cast of the Google Factory Tour, researcher Franz Och presented the current state of the Google Machine Translation Systems. He compared translations of the current Google translator, and the status quo of the Google Research Lab’s activities. The results were highly impressive. A sentence in Arabic which is now being translated to a nonsensical “Alpine white new presence tape registered for coffee confirms Laden” is now in the Research Labs being translated to “The White House Confirmed the Existence of a New Bin Laden Tape.”

How do they do that? It’s certainly complex to program such a system, but the underlying principle is easy – so easy in fact that the researchers working on this enabled the system to translate from Chinese to English without any researcher being able to speak Chinese. To the translation system, any language is treated the same, and there is no manually created rule-set of grammar, metaphors and such. Instead, the system is learning from existing human translations. All it needs is someone to feed the system the two books and to teach it the two are translations from language A to language B, and the translator can create what Franz Och called a “language model.” This is brute force AI, if you want – it works on statistical learning theory only and has not much real “understanding” of anything but patterns.

Now comes the exciting part. This translation system can have numerous applications. Apart from the usual 'website translation' services, it could be used in more advanced applications like Instant Messengers and Babelfish-It would be a smart device you plug-in to your ear, and it would have speech recognition and Auto-translation built in. You can now visit a foreign country and understand people who talk to you in languages you never learned. While the needed text-to-speech and speech-to-text technologies are far from perfect at the moment, they are still realistic possibilities.

No comments: