OmniPage Pro - Oct/Nov '98

OmniPage Pro

When shopping for an optical character recognition (OCR) package to translate your hard-copy documents into editable text, interface and ease-of-use are important, but the three foremost factors in selecting an OCR application are accuracy, accuracy and -- you guessed it! -- accuracy.

Most OCR programs produce 95 percent accuracy from good originals, but the true test is how they process poor originals, skewed documents, faxes, and second- or third-generation copies. Caere Corporation's immensely popular OmniPage Pro, now in version 8.0 for the Macintosh as well as the PC, touts a new recognition engine designed to accurately scan these notoriously tough documents.

Caere completely revamped OmniPage's recognition algorithms in version 8.0 to move closer to 100 percent accuracy for poorly formed documents. My tests indicated that Pro 8.0 is indeed more accurate, producing surprisingly good results for even the most difficult of documents.

I tested OmniPage 8.0 for Mac on an Apple Macintosh 9600/300 and compared speed, accuracy and throughput with OmniPage 6.0 on a 9500/200. The results for version 8.0 were impressive. On a clean original with 12-point Courier type, I achieved 100 percent accuracy almost every time with both versions. But on difficult documents, OmniPage 8.0 Pro came out ahead every time. Using an incredibly difficult to recognize test image of 4-point type with two columns and skewed approximately 5 percent, Omnipage Pro 8.0 -- using its new Deskew feature, which re-orients skewed documents up to 10 percent -- produced acceptable results (I didn't have to retype the image), while version 6.0 resulted in gibberish.

8.0 is also faster than previous versions, particularly in throughput speed (i.e., how long it takes the program to produce editable text from scanning through proofing).

OmniPage's OCR proofreader has also been enhanced in 8.0. Formerly called "Check Recognition," the proofreader verifies recognized text against the scanned image by displaying a bit-map of the suspicious text during the proofing process. You compare the bit-map with the suggested word and either make the correction manually or select an alternative word from a list offered by the program. Questionable words or characters are tagged by colors to identify them as suspect. The program can also be configured to disregard abbreviations, proper nouns or other unusual words. Overall, the proofreader significantly improves the time it takes to go from hard copy to editable text.

OmniPage 8.0 remains a highly intuitive program to set up and use. It installs using a CD-ROM containing program files and various drivers for scanners.

OmniPage supports most image scanners from Apple, Hewlett Packard, Umax, Microtek and others. If your scanner shipped with a TWAIN data source, OmniPage Pro will use the driver to access the scanner. Caere also provides support for the latest scanner-related files from their Web site at www.caere.com.

OmniPage 8.0 includes some substantial automatic input/output system controls. For instance, you can schedule OCR for documents while the you're away from the computer. If your firm receives a document that must be OCR'd, such as a series of fax files, simply save the scanned or faxed file to a specific folder and OmniPage will regularly access the file for pending documents and perform the OCR function unattended.

With so much to commend, what are my complaints? OmniPage does some strange things at times. For instance, it wouldn't save a OCR'd document in any text format on one occasion. In fact, none of my save commands in any program worked until I rebooted my computer and lost the processed text. I was unable to recreate the bug in future sessions, however.

On another occasion, the program produced a large number of tabs in a processed document that required some time to clean up on WordPerfect. I believe, however, that a proper setting of preferences would eliminate this problem. At other times, I couldn't cancel out of a dialog box that attempted to save an incorrectly spelled word to a user dictionary.

These were minor annoyances, however. All in all, I was impressed with the new version of OmniPage Pro. You may be too.

Caere Corporation
(408) 395-7000
www.caere.com
Windows 3.x, 95, NT,
Macintosh

$499

Reviewed by David A. Saraceno, sole practitioner and consultant, Coeur d'Alene, Idaho

Oct/Nov '98 Issue

PROS
Quick and accurate.

CONS
Somewhat quirky.

VERDICT
Can OCR almost anything.