The future of OCR
Dec. 20th, 2010 01:57 am![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
nancylebovOne of my commenters suggested that old books will never be properly OCRed because the pages are apt to be warped because of early paper-making.
I believe this is a temporary problem-- that degree of warping is just mild and fairly simple geometrical transformation.
[Poll #1658707]
If the usual captchas are useless, one solution is to go to simple semantic captchas. I've seen a few of those which require answering a very simple arithmetic problem or figuring that "What is the opposite of hot?" isn't a trick question.
Are there other possibilities? Will the reverse Turing test (machines recognizing people) continue to be reliably solvable?
I believe this is a temporary problem-- that degree of warping is just mild and fairly simple geometrical transformation.
[Poll #1658707]
If the usual captchas are useless, one solution is to go to simple semantic captchas. I've seen a few of those which require answering a very simple arithmetic problem or figuring that "What is the opposite of hot?" isn't a trick question.
Are there other possibilities? Will the reverse Turing test (machines recognizing people) continue to be reliably solvable?
