You know those squiggly little word puzzles you have to solve to prove you’re a flesh and blood sentient being on many sites? If the site was using reCAPTCHA, there’s a good chance that you were helping to translate old books, newspapers, or radio shows.
reCAPTCHA is helping to digitize the massive catalog at Google Books and also old editions of The New York Times. I know what you’re thinking. Aren’t computers just scanning and “reading” the material? Well, yes, but Optical Character Recognition (OCR) isn’t quite perfect yet, so it doesn’t always recognize a word. And that’s where we all come in. Some jobs are best done by humans, and figuring out a blurry word is one of them.
So, how does the computer know we’re right? When we’re prompted to solve a CAPTCHA, we’re presented with a word pair. One word is known, the other is not. If you get the known one right, it’s assumed you probably got the other correct, and so you’ve just translated a word for all of posterity. Pretty cool, eh?
Building on this concept, Duolingo, a company started by CAPTCHA inventor Luis von Ahn, will teach you a foreign language by having you do ever increasingly complex translations from your native tongue into your target language. The crowdsourcing goal for this one? Oh nothing too heavy, just cross-translating THE ENTIRE WEB into every language.