Is reCAPTCHA's data used to improve any OCR software?

reCAPTCHA is used in lots of places and no doubt has a massive amount of data of 'bad l开发者_如何转开发ooking image' to 'word'. This sort of data could be used to make really good OCR software. So is there any OCR software that is using this data?

Possibly as a training data, I doubt there's a way to find out. But basically, it is you to be used as an OCR software (your input - if confirmed - is used to decode OCR-unreadable words in scanned books).

doubt it. i was gonna answer that if google would release data associated with recaptcha it probably would turn against them. but in reality anyony can train it's own neural network to do what google does; the problem is, that many images served by recaptcha are in so bad quality, that taking any conclusions (or traning naural networks) from this images would produce very high false-positive results