What do you think of this fact?
Remember those warped words you had to type to prove you weren’t a robot? One of them was a test. The other was real work for a massive digitization project.
An idea born on campus
reCAPTCHA was created by a team at Carnegie Mellon University that included computer scientist Luis von Ahn. He had already helped develop the original CAPTCHA. Then he realized people around the world were burning billions of hours typing those things. His idea: if everyone is deciphering words anyway, why not make them words someone actually needs?
When old texts get scanned, text recognition software often chokes on faded or smudged print. According to the Smithsonian, the software missed 10 to 30 percent of nineteenth-century English text. Those problem words were exactly what ended up in the CAPTCHA.
How the system put you to work
Each reCAPTCHA showed two words: a control word with a known answer and an unknown one from a scan. If you got the control word right, the system trusted your answer for the other one too. Once several people typed the same thing, the word was cracked.
The pilot project was the New York Times archive. In a test with 24,000 words from Times archives going back to 1860, reCAPTCHA hit 99.1 percent accuracy, while standard text recognition managed only 83.5 percent. In the first year, users solved more than 1.2 billion reCAPTCHAs and deciphered over 440 million words. That’s roughly 17,600 books. Later, scans from Google Books joined the mix.
Google bought reCAPTCHA in 2009. Around 2012, the book words were swapped for Google Street View photos. That’s where all the traffic lights and crosswalks came from.
And then came Duolingo
Von Ahn stuck with his big idea: lots of people together can solve what computers can’t do alone. He later co-founded the language app Duolingo with Severin Hacker.
So if you ever cursed at a CAPTCHA, take comfort. You helped rescue a piece of newspaper history.