9 Mar
1999
9 Mar
'99
12:29 p.m.
That's not what I'd call "high".
That means that on average, you have to
correct or interpret every tenth character. I'd call less than 99% "low",
not high. Our Department looked at this a few years ago, and rejected
anything less than 95%, I think. Even that means correcting (or as one
person put it, "clicking on") one character in every twenty.
The best solution for this is to keep the scans AND the OCR'd text. That
way, with a simple database, one could do searches on the text, and get
most of the hits, yet actually read the images.
William Donzelli
william(a)ans.net