Classic Computer Magazine Archive COMPUTE! ISSUE 132 / AUGUST 1991 / PAGE 126

Typist. (hand scanner) (evaluation)
by Robert Bixby

At first glance, you might think Typist is simply a hand scanner. It's little wider than a standard hand scanner, but not enough to make much of a difference. But Typist's OCR (Optical Character Recognition) software makes it noteworthy.

The scanner requires at least 2400K of free extended memory in an 80386 or 80486 system. That means you'll need 4MB of RAM to operate Typist. This device comes with a short 8-bit card that takes up a slot in your bus. The installation program takes care of itself, and the manual is very explicit, providing adequate hand holding to comfort all but the rankest neophyte. Once you've installed Typicst, you can use it for capturing either images or text. Generate graphics from photographs or line art and then store them in bitmap format (TIF or PCX). You must set the scanner for monochrome with a switch on the body of the hand scanner to use the OCR software, but you can use either monochrome or one of three diffferent levels of dithering for grabbing graphic images.

Typist's OCR software component works in the background as a TSR in DOS or a minimized application in Windows. As soon as you press the button on the hand scanner, you activate the software, which buffers the bitmap image and interprets the characters. Once the OCR has figured out all of the letters it can from the scanned image of the printed page, it outputs the information to the keyboard buffer, where (to your application) it looks exactly like text you're typing in.

One of the nicest aspects of the OCR software, since even this slightly wider hand scanner can't quite span the width of a page, is that you can scan sideways, top to bottom or bottom to top. Just set the OCR software to automatically read the text in the proper direction. Or set it to read text in only one of these directions. If you're scanning text in a columnar format, you can set the OCR to pay attention only to the first, middle, or last column of text. And, as if that were not enough flexibility, the system can zip scans together so you can scan a wide page of text in a series of horizontal bands and let Typist automatically detect the overlap and eliminate the extral lines. Usually.

As your reward for reading this far, here's the Kitty Kelley lowdown on OCR: The truth about OCR software is that Caere and a couple of others are at the forefront of a fledgling technology. One of the reasons Typist demands so much memory is that identifying text characters is next to impossible, particularly when you're scanning a variety of typefaces. Typist, and almost every other OCR system, falls to pieces when it runs across italic text (or any other type style slightly out of the ordinary). It can't make out one italic letter in ten. Likewise, if your contrast setting is too dark or too light, your scan is slightly unsteady, or the scanner isn't perfectly straight with the page, Typist will make a lot of errors.

Despite this, I was amazed at the power of Typist. It performed as well as a flat-bed scanner on average, and its software seems to be slightly ahead of ReadRight in terms of power (it even reads dot-matrix printouts without a hitch), and it's much friendlier. The price of the Typist unit puts a flat-bed scanner and dedicated OCR package to shame. In spite of all the editing scanned text requires, it's miles ahead of typing in text in terms of speed and convenience.