3.789 scanning (49)

Willard McCarty (MCCARTY@vm.epas.utoronto.ca)
Mon, 27 Nov 89 22:02:02 EST

Humanist Discussion Group, Vol. 3, No. 789. Monday, 27 Nov 1989.

Date: Mon, 27 Nov 89 12:07:10 PST
Subject: [DCGQAL]A0234!On Scanning and OCR-ing

On Scanning and OCR-ing

My own opinion is that the current level of accuracy to be expected from
OCR scanning is less than satisfactory for most scholarly uses. The
Kurzweil scanner (top of the line) in a project with which I was
involved a couple years ago, generated something on the order of 2-3
errors for every 50,000 characters, (equivalent to about 2-3 errors per
"galley" in traditional typesetting terminology.) In that case, and in
subsequent cases, I've found it more accurate (read: cost-effective) to
rekeyboard the texts and have them delivered in machine-readable form on
floppy diskettes. I've used an outside vendor with considerable
satisfaction, who manages an off-shore division for just such purposes.

For a price and additional information you may wish to contact:

Gordon Johnson
Interactive Composition Corporation
2255 Morello Blvd
Pleasant Hill, California 94523
(415) 686-6860

Just in anticipation of reactions to sending work off-shore (since
objections have been voiced before), I've always felt there was a
positive good in being able to export decent work which could
more-economically be produced in an offshore economy than in our own.
For too long, we've tended to dump our surplus products to Third World
countries irrespective of need or value Providing humane and satisfying
job opportunities that do not endanger local health or ecology, seems a
far more responsible economic interaction. Certainly in the case of
keyboarding, the work needed can be accomplished with considerable
satisfaction and profit to all parties.

Chet Grycz
Scholarship and Technology Study Project
University of California