Spurred on by the announcement today of the purchase of reCAPTCHA by Google, I did some belated research on how CAPTCHA works.

Its primary goal is to prevent the filling in of online forms by nasty BOTS. The idea being that an automated BOT will be unable to recognise and decode the murky letters presented by a CAPTCHA form, while a human will usually be able to decode the words. Usually that is, because I am sure that like me, you often fail to get the decode part right first time. The CAPTCHA will then force you to keep trying until you get it right, which can be a major source of irritation.

The thing is that the irritation is worth it because it keeps all but the most sophisticated BOTS at bay.

By way of an example, I had customer with a simple form online input form for sales enquiries that did not have a CAPTCHA on it. The result was that the BOTS filled in the form at a rate of 4 to every 1 human entry.  This was a major pain in the derriere because each day someone has to sift through the submissions to sort the wind from the chaff.

As soon as we added CAPTHCA to the form, BOT submissions dropped to ZERO.

Brilliance of reCAPTCHA

The reCAPTCHA organization is involved with the digitizing of old text from various sources such as books, and newspapers. The problem is that some of the old text is too aged and blurry to be deciphered by Optical Scanning Recognition (OCR) software. This prevents the process from being automated and this is where we come in.

If you look at the reCAPTCHA example above you will see that two images are presented for the human viewer to decode.  What happens is that the computer knows what one image says but is clueless about what the other says! So it presents them both to you and waits for your answer.

If you get the right answer for the one the computer knows, it automatically gives you 100%. By doing this, it assumes that your other answer is also correct and then lets you submit the form or access the page.

The smart part is that the computer stores all the answers to the word it does not know and eventually works out what the correct word is based on a consensus of answers from a number of human responders. The image of that word then gets marked as deciphered and a new image gets put up for deciphering.
In this way with the 200 million average reCAPTCHA tries each day, they are slowly digitizing various old records. At this time reCAPTCHA says that they are busy with old editions of the New York Times.

And you can be proud because you are helping to get the job done!