reCAPTCHA - neat idea

[drwex]

A CAPTCHA is a supposed test to tell if a form is being filled out by a human or a bot. It presents a distorted image and asks you to type in the original word, answer a question, or solve a simple math problem. CAPTCHAs are often used to try and stop spam bots.

CMU has come up with an interesting twist on the technique. reCAPTCHA presents words that failed OCR, not made up images. The puzzle answers supplied by humans are then used to improve these digitized books so they can be better searched, or more accurately computer-read to blind people. Currently reCAPTCHA is working with The Internet Archive, which I consider a worthy endeavor.

CAPTCHAs themselves have some issues, such as reducing accessability. But if you're going to use this technology, reCAPTCHA seems to be a nice way to go about it.

Comments

[woodwardiocom.livejournal.com]

reCAPTCHA presents words that failed OCR, not made up images.

If it failed OCR, how do you know what the right answer is?

[ckd]

They give you two words; one's the OCR failure and the other is a known word. You don't know which is which.

[woodwardiocom.livejournal.com]

Gotcha.

[drwex]

In addition they present the same OCR-failed word to several people and correlate the answers.

[qwrrty.livejournal.com]

That is cool!

I would be worried that presenting the words out of context will make it more difficult even for humans to tell what word it should actually be. When the word is unclear, seeing it in the proper context can go a long way toward figuring out what was meant. But most words that were unclear for OCR will probably still be easy for people to get, even out of context.

[cos]

Oh, so meta! See, this is very close to the most clever way of *defeating* captcha these days: Run some porn sites that require viewers to solve captchas to see the images. Instead of using/generating your own images, collect captcha's off the web on the other end, and feed those to your viewers... then feed the solutions back. So on one site, you have a porn site that gets humans to solve captchas; on the other side, you have a spam engine that solves captchas by feeding them back to the porn site and getting human answers back, which it uses for the web sites it wants to get access to.

These two systems could tie together unknowingly. Somewhere out there, someone trying to view porn may be feeding a captcha solution to a spam engine which in turn is feeding that solution back to a recaptcha system that is using it to digitize books.

The Internet is so much greater than the sum of its parts.

[qwrrty.livejournal.com]

These two systems could tie together unknowingly. Somewhere out there, someone trying to view porn may be feeding a captcha solution to a spam engine which in turn is feeding that solution back to a recaptcha system that is using it to digitize books.

I love it! This image is going to be bouncing around in my head all day.

[rdhdsnippet.livejournal.com]

I hate those damn things. I get them wrong at least 75% of the time.

[intuition-ist.livejournal.com]

grr. i really like your journal, but this (http://desert-born.livejournal.com/132066.html) still applies.

:(

[drwex]

*shrug* The "use this capability" basically means I'm marking my LJ as potentially containing things discussed by adults. You can make this all go away by setting your birth age in your profile to some date that makes you appear over 14 or 21 or something.

If you don't wish to do that, I understand the choice.

[intuition-ist.livejournal.com]

i understand why you want to use the capability -- i'm saying the capability itself is ridiculous. "adult" is the norm and the majority. i should not have to "prove" (in quotes for the utter ridiculousness of it, since the "proof" currently consists of clicking a button, which any 3-year-old could do) i'm over 14 or 18 or whatever to read anyone's journal.

it's a preposterous solution to a problem created by technology and lawyers.

[intuition-ist.livejournal.com]

in other words -- the emperor has no clothes, and all the cut tags in the world won't change that.

[drwex]

I don't disagree with you in the least. I suppose it's just that I don't expect LJ to be any better so I'm neither shocked nor disappointed when they act this way.