Scanning a picture saves typing 1,000 words
I REMEMBER my first fax. It was 1978. As a summer intern at a news-wire service in Chicago, I was in charge of setting up the fax machine whenever the New York office wanted to send a document.
Faxes were big deals back then. I had to take a cylinder out of the machine, wrap a sheet of paper around it, then pop the cylinder back in. When New York called, I would stick the telephone receiver into rubber couplers. The machine would beep and whir, then send the cylinder spinning furiously. Several minutes later, I had a one-page fax.
Fascinating stuff. But cumbersome. Which brings me to computer scanning. A scanner works like a reverse copier. It takes a picture of a document and puts it on a computer's hard disk. You can store that image, which takes a lot of disk space. Or, if it's a text document, you can use optical-character-recognition (OCR) software to turn the picture into a much smaller, word-processor-ready text file.
Like the fax machine, OCR has come a long way. When I tested it
2-1/2 years ago, it was slow and tedious. Today, it's nearly fast enough and reliable enough to become a mainstream product.
The signs already point that way. OCR is popping up in computer fax programs and as a free accessory with scanners. As scanners get better and cheaper (desktop models are selling at roughly half their 1992 price), OCR software sales are rising. Kristy Holch of BIS Strategic Decisions predicts the scanner market will grow 25 percent annually in the next several years.
But don't jump into OCR until you've got the right computer. You need at least eight megabytes of random-access memory (RAM) and a really fast machine to make the software work comfortably. During my 1991 tests, the 386-class computer with four megabytes of RAM choked on Caere's OmniPage Professional program. On my new 90-MHz Pentium with 16 megabytes of RAM, the updated OmniPage Pro 5.0 screams. So does Calera's new WordScan Plus 3.0. I've gotten the Pentium to convert an 80-page booklet into text about as fast as I could scan it into the machine.
Many users make do with far less powerful machines. (The Macintosh version of OmniPage doesn't even work with the fast PowerPC models.) But I doubt OCR will really become mainstream until it's as easy and fast as faxing. Calera's president thinks that in three to five years OCR will be a simple button you click in a word-processing program.
WordScan and OmniPage are generally considered the best OCR programs for personal computer users. But they still stumble a bit. The programs allow you to proofread the text - a necessary step.
I tested both programs on three scanned documents: 800 words from the Wall Street Journal, a short press release, and a single page of ``The Grapes of Wrath.'' WordScan did best with newspaper text. It missed a whole line of type (``Telecommunications Policy'' became ``T nuin 9 ftky'') but ended up making only 14 mistakes to OmniPage's 23. On the other two tests, OmniPage scored better: three mistakes versus 18 on the 300-word press release (which included tiny blue type) and only one mistake versus five on 250 words from ``The Grapes of Wrath.''
Overall, I preferred OmniPage. Its layout was simple enough that I didn't have to open the manual. It did what it said it would do out of the box. WordScan Plus was far less intuitive. The result is that I'm scanning more than ever, retyping less and less, and wrapping fax cylinders not at all.
* Comments? Send them to: CompuServe (70541,3654), Prodigy (BXGN44A), or Internet (email@example.com).