Ocular Character Recognition


Ever since individual personal computers first came on-line in large numbers, they have been utilised as a huge opt-in distributed computing array by projects such as SETI at Home and Folding at Home.

But there are information processing tasks that can be distributed yet are still impossible to perform with computers. The Stardust at Home project uses the unparalleled image-recognition capabilities of the human brain to process data from an interplanetary sample collection mission. People all around the world take part in their spare time.

Auntie Beeb's weekly program on the worldwide use of digital tech, Digital Planet, now reports on another innovative scheme to harness the eyes and noggins of computer users.

You know when a web site displays squiggly text against a blurry background and asks you to type in the characters to prove that you are not a spambot? The CAPTCHA project at Carnegie-Mellon uses this to correct the character recognition of scanned old books.

Optical Character Recognition (OCR) is pretty good these days, partly because the algorithms now use dictionaries and word-frequency tables to improve their guesswork. But the technology is still far more error-prone than a human reader. And OCR can apparently identify which words it's having problems with. So the CAPTCHA project does something very smart.

When you need to pass a spambot test at a participating website, the project's server feeds you two words from books it's working on. One is a word it knows. One is a word it's having trouble identifying. It doesn't tell you which is which. You type in both words to gain access to the web site in question. The server thus collects a number of interpretations of each tricky word, and when a certain interpretation gets enough human "votes", it is accepted as correct. Beautiful! People around the net take hundreds of thousands of these tests every day. Instead of asking people to devote spare time to the project, like Stardust at Home, the CAPTCHA project uses brain time that would otherwise just go to waste.

[More blog entries about , , , ; , , , ]


More like this

In this post: the large version of the Brain & Behavior and Technology channel photos, comments from readers, and the best posts of the week. Technology. Reflections from the nose cone of a propeller plane. From Flickr, by Katie@! Brain & Behavior. From Flickr, by jamesfischer Reader…
What would you do if someone asked you to help transcribe an old book onto a website? Chances are, you'd say no on the basis that you have other things to do, or simply that it just doesn't sound very interesting. And yet, millions of people every day are helping with precisely this task, and most…
Traveling and busy as hell, but wanted to share this. The ever expanding copyright laws is one of my pet peeves, but almost as irritating as the increasing length of copyright is the difficulty in knowing if something is still under copyright. The copyright date and name of the copyright holder in…
Human eyes and brains are still way, way better at image recognition than computers. There are many visual tasks that we do swiftly ourselves but that we can't yet get machines to do reliably at all. In January of '06 I blogged about the Stardust @ Home project where you can help identify…

That's one of the tidiest little bit of technology I've read about this week.

Yeah, that's incredibly neat. But kind of old news; they have been doing that for what, more than a year by now? I'm surprised no major tech news outlets have reported on it earlier. Or have they?

By Niklas Ramsberg (not verified) on 20 Aug 2008 #permalink

OCR-scanning is the greatest invention since sliced bread. Now if they can just create smarter translating programs my life would be bliss.

Increadibly smart this CAPTCHA project. I had heard something vague about it but didn't realize it simply utilised this common security measure. I had started wondering why real words had started to appear instead of just random letters and thought it was meant to make it easier for users. Well, one kan kill two birds with a stone I guess.

I just hope they don't get sued for salaries by greedy users.

...maybe it just means it expects the reader to be female? ;-p

Seriously, I did not know that the CAPTCHA tech was actually doing this. That is damn clever.

By Luna_the_cat (not verified) on 29 Aug 2008 #permalink

Dear sir,
I need the full source code for captcha image in asp.net with vb.net 2.0.Please help me.

Hi Martin..
I am trying to recognize the alphabet in captcha using template matching ..but still i couldn't find.
Could u plz tell me some method to identify the character

By Sakthi Gs (not verified) on 06 May 2009 #permalink

dear sir,

we want this project pls send this project detail's in my email id.

By chokkamsivaprasad (not verified) on 15 Oct 2009 #permalink