New site to use crowd-sourcing as means to translate the internet

After figuring out a way to make good use of those annoying CAPTCHA's, by using scanned text from old books to create digitized versions (reCAPTCHA) while still just a grad student, and then selling the results to Google, and then following that up by figuring out a way to get people to label pictures for free to improve image searches (Google Image Labeler), Ahn is now set to apply his impressive crowd-sourcing skills to the areas of both language learning and translation. He believes Duolingo will attract some portion of the estimated billion people who are interested in learning a new language, because unlike other good language-learning sites, Duolingo will be free.

Working with PhD student Severin Hacker, for the past year and a half, the pair struggled with ideas on ways to convince millions of people to translate language text, for free, until hitting on the idea that it would have to be associated with something they’d want to do without prodding; in this case, learn a new language. And then, like Wikipedia, the results of the amateur’s translation process as they learned, would be honed as others worked on the same piece until the text reached some point of maturity, whereby it would be released to the web.

Ahn claims the results of his crowd-sourced translation process creates text that is just as good as a professional service, which means that if history is any kind of gauge, sometime in the near future, when you use Google’s translation feature on a foreign language site, the results will likely be far superior to what you get currently when a machine does it for you.

© 2010 PhysOrg.com