http://recaptcha.net/learnmore.htmlHow fascinating ... books are getting digitized one word at a time by Internet users! You've seen those images with distorted text pop up on blogs or websites. You type in the word you see in the image, and then you can leave a comment or send email to someone. If the word is not typed in correctly, then access is not given. The test is done to weed out robot programs that are harvesting email addresses for spammers, or putting unwanted advertising comments in blogs.
This project by Carnegie Mellon is using scanned text from digitization projects in those test images, and the resulting text typed in by human users is fed back to the digitization project until the whole project is completed.
I first read this and couldn't figure out how it worked. Don't you need to know the person typed in the right word before giving them access to leave comments or email? So I read up on the details. They actually have the user type in two words, one word they do know, and one they don't. Access is given on based on the known word, and the translation of the unknown word is noted. They feed the same unknown word to several people. If all of them identify the unknown word with the same word, then the software figures the people got it right and feeds their translation in the digitization project.
Are any of you using reCAPTCHA?