Thanks for the shout. I've got my current project in the box. Both "gocr" and "ocrad" had me correcting 30-40 errors per page. Neither seemed to be overly concerend with run-together words. Perhaps that is a problem with how the dictionary is used by those programs. I ran the ocr results through Open Office one more time and that seemd to clean up most of the stuff. You'd never be able be able to use these programs to do automated forms processing. It'd require too much operator intervention and probably be just as easy to re-transcribe by hand. In a production setting, I'd probably be able to justify the purchace of well developed software ... just now, I'm only playing. Thanks again, Harv On 10/24/05, Leif Johnson <leif.t.johnson at gmail.com> wrote: > > > > > I've tried "gocr" and "ocrad" ... the results are less than favorable. > > I'm looking for something with an error rate of less than 10 per 1000 > > charaters. > > I don't think you will find it. OCR that is this good is usually doing > target matching, which doesn't sound like it is an option for you. What kind > of error rate are you getting anyway? You might just have to bite the bullet > and do a little keying for low confidence results.... > > leif > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://shadowknight.real-time.com/pipermail/tclug-list/attachments/20051024/5d81a0cb/attachment.htm