Software that can read - Software Review - Caere's OmniPage Professional and Xerox Imaging System's TextBridge OCR software - Evaluation

Home Office Computing, Sept, 1994 by Russell Letson

Rating: ***1/2

OmniPage Professional

Version Reviewed: 5.0

List Price: $695

Average Street Price: $485

Publisher: Caere, (408) 395-7000, (800) 535-7226

WIN

Rating: ***1/2

TextBridge

Version Reviewed: 2.0

List Price: $99

Average Street Price: $79

Publisher: Xerox Imaging Systems, (508) 977-2000, (800) 248-6550

WIN / MAC

Even with all the talk about the paperless office, getting buried in paperwork is still a common problem. Storing faxes, magazine articles, or other hard copy materials on your hard disk rather than in a file folder makes them much more convenient to access, manipulate, and organize. With today's optical character recognition (OCR) software and a scanner, you can easily transfer the contents of these pages to your computer where, as electronic documents, they can be imported into your favorite word processing, desktop publishing, or spreadsheet applications and be edited or searched to your satisfaction.

OmniPage Professional 5.0 and TextBridge 2.0 display the accuracy and range of functions possible in current-generation OCR software. Both programs function effortlessly with a variety of specific scanners (as well as with any scanner that supports the TWAIN specification), read a host of typefaces and sizes, and export text to major word processors, spreadsheets, and desktop publishing programs.

We ran a variety of materials, including an elaborately laid-out magazine article and a second-generation photocopy of a page of type, through both OmniPage and TextBridge. OmniPage handled the magazine pages well, keeping track of the columns and font sizes, and even recognizing a drop cap initial letter. (It did, however, repeat part of the line next to that drop cap.) The blurry photocopy presented the biggest, and probably most common, challenge, but suprisingly OmniPage and TextBridge got through it with only half a dozen errors.

OmniPage is the more expensive and complex package, aimed at the high-end market: people who work with a variety of document formats and need to save not just text but original page layouts, fonts, and graphics. You might, for example, need to edit and update a manually produced catalog or price list, but you don't want to discard your previous work and start from scratch. When you scan the work using OmniPage's True Page mode, it not only reads the text, it retains the font and layout--including column and paragraph as well as graphics--so when the text is exported to a supported word processor or desktop publishing application, minimal editing is needed to restore it to its original form.

OmniPage employs a battery of techniques (help-fully explained in the clear, manageable, and extensive manual) to read difficult areas such as fuzzy, faded, or blurry copy and to identify text zones--those places on the page where text belongs together, such as in columns or header blocks. It also uses its rule-and dictionary-based Language Analyst to figure out ambiguous words or letters. Its editing and proofing functions then prompt you, if necessary, to check and hand correct its guesses on difficult cases.

All this demands plenty of computer resources, however: A system with 8MB of RAM is considered minimally configured (12MB is preferred), and OmniPage needs lots of hard-disk space (18MB) for the program and its temporary working files. What OmniPage delivers for all that memory outlay, however, is quite impressive. Its operations are well laid out, extensively customizable, and capable of working in automated or step-at-a-time modes. The five manuals (including one for the Image Assistant graphics editor) are clear, copiously illustrated, and indexed.

This program is not without rough spots. For example, the True Page mode (the feature that allows the program to retain complex layouts in a scanned document) is incompatible with WordPerfect for Windows 5.2. We set OmniPage to Ignore Fonts and All Formatting, but some text still showed up in italics in addition to all sorts of embedded formatting information that took a fair amount of hand editing to remove. The only way we were able to get plain output for WordPerfect was to save the text in the ASCII format.

TextBridge is an easy-to-use and very affordable alternative to OmniPage. Although it lacks the elaborate format, font, and graphics retention features of its sophisticated counterparts, it's anything but second rate at the basic task of text recognition. (We tested version 2.0 for Windows, but version 2.0 for the Mac is functionally the same.) TextBridge handled both block-format and multicolumn pages without a hitch (even though it didn't read the drop cap, it went on to get the rest of the line right) and also performed well in the furry photocopy test, producing a usable file that needed about the same amount of hand correcting as the OmniPage version.

TextBridge is more modest in its demands on system resources, but it is quite flexible, perhaps better suited to a non-OCR-intensive working environment. One of the most useful features of TextBridge is its application server, which allows the program to operate (with limited functionality) from within other programs. With TextBridge running in the background, the program will automatically appear in the file menu of any word processing or desktop publishing program that you open. This allows for complete ad-hoc document recognition as well as for insertion of scanned material while you write.