CAPTCHA technology improves OCR, which is ultimately good for everyone.
Google today acquired reCAPTCHA, a tiny firm originally started as a project of the School of Computer Science at Carnegie Mellon University. CAPTCHAs are those funny little stretched out text images used with authentication. They can’t be read by a computer, so they test that a human being is trying to access an account, and not a bot. But oddly enough, it is because computers can’t read them that the company reCAPTCHA has found a way to use CAPTCHAs to improve the accuracy of optical character recognition.
OCR is the only practical way to scan large volumes of printed materials into an online, searchable archive. But OCR is, and always has been, awful. It misinterprets a large percentage of characters, creating muckity-muck out of what was once clean text. And if it is your goal to scan every book in the world, even orphaned books, as is the stated purpose of the controversial Google Books projects, then better OCR is an urgent matter.
But how can CAPTCHAs improve OCR? Here’s the cool part of what reCAPTCHA has been doing. Because about 200 million CAPTCHAs are solved by humans worldwide daily, reCAPTCHA says, it can tap into the power of all those human minds to solve the OCR muckity-muck. It presents people with a multi-word CAPTCHA for each login attempt. At least one of the words is a known word, the other(s) is from an OCR scan that produced a wrong, misspelled word. If the person correctly reads the known CAPTCHA word, the system assumes the person correctly read the other word, too. It cross-tests this word by sending it out to many people, who all read it and tell the computer what it is and hence they clean up the misspelled words produced by the OCR scan.
Smart stuff. Google has the scale of users to put this to work for Google Books and its Google News Archive Search. This doesn’t mean that either of these projects are themselves good ideas. I still side with the folks that oppose Google Books (at least as it is currently set up). The Google News Archive is another can of worms with similar issues. But such a large-scale use of reCAPTCHA to improve OCR could be the silver lining that at least partially redeems Google’s otherwise borderline ideas.
reCAPTCHA is only the second acquisition that Google has made in 2009, bought for an undisclosed amount. In August, Google gained video compression technology when it purchased On2 Technologies for $106.5 million.
Like this post? Visit the Google Subnet home page for more news, blogs and podcasts. Sign up for the weekly Google newsletter. (Click on News/Google News Alert.) More blog posts from Google Subnet:
- My, What a Long Tail You Have
- Searching For Cream: The Endless Battle To Find The Best
- Google is not interested in competing against Apple
- Two startups offer free ‘computers’ as cloud services
- Amazon’s latest competition offers secure enterprise cloud
- Idea Checklists: How Companies Can Utilize the Power of Social Media
Subscribe to all Google Subnet bloggers or Follow Google Subnet on Twitter
Follow Julie Bort on Twitter.
