[Gardeners] [Lisp] SoC Idea: Optical Characters Recognition With Learning Capabilities

[Stuart Sierra]

Hedos wrote:
> Hi folks, I have an idea for the Summer of Code, I would like to know
> what you think about it and see if any possible mentor might be
> interested.
> 
> This is by no mean a formal proposal at the moment.
> 
> ** Project title **
> Optical Characters Recognition With Learning Capabilities
> or
> Human Trainable CAPTCHA Solver

Solving CAPTCHAs is an interesting test of OCR capabilities, but I for 
one would be far more interested in a *useful* open-source OCR 
application (better than JOCR) that can learn to recognize different 
fonts and layouts, ignore images, and work with poor-quality photocopies.

Of course, an OCR that can solve a tricky CAPTCHA might be pretty good 
at general-purpose OCR as well.  But working with full-length documents 
might allow it to make guesses based on the content of the text as well 
as the image.

-Stuart