Xerox Imaging Systems has released a Mac version of its highly regarded TextBridge optical character recognition (OCR) software, available for more than a year under Windows.
It’s a bargain — the current market leaders in OCR cost $500 — but that recommendation includes several caveats. On a set of three exceptionally clean documents (in 14-point Times Roman printed at 600 dpi), TextBridge scored 100 percent accuracy. That puts it on a par with Caere’s OmniPage Professional, a Macworld Editors’ Choice in our last OCR roundup (November 1993). TextBridge also turned in a fine performance (95 percent accuracy) on the body text of a scanned paper fax, a particularly challenging document since the text resolution had effectively been degraded twice. In competition with the six Mac OCR-roundup packages on a variety of documents, TextBridge would probably place no lower than second. Accuracy on italic and bold text was exceptionally good.
TextBridge’s recognition problems are truly baffling. On a document from a statistics-software vendor, it got hierarchical character analysis right, but read the as tne all the way through the text. Xerox’s Lexifier technology is supposed to guard against this and other low-level errors. To its credit, TextBridge’s Verifier mode enables you to train the program to recognize easily misconstrued words in a particular font; at the expense of considerable processing time (several minutes per page on a Power Macintosh 7100), you can push reading accuracy to 98-plus percent on nearly any font at point sizes larger than 6, if you’re willing to verify 5 to 15 pages of text.
TextBridge has settings for text quality, previewing, verifying, image orientation, zoning, and language character set (English, French, German, Italian, and Spanish). The Auto Zone feature correctly chose columns in every multicolumn document I tested, although blocks of line art usually confused it. TextBridge supports every currently available scanner and the TWAIN source manager, and it can process scanned documents to sets of TIFF or PICT files for later recognition, but it could not read TIFF files saved in Photoshop 2.5.
AppleScript gives this product extra power for heavy-duty use, in one specific way. It’s possible to train TextBridge on your documents (to attain acceptable accuracy) and fine-tune some of the AppleScript examples to recognize and store massive sets of files automatically. TextBridge is priced as an entry-level product, but its AppleScript capability is likely to appeal mostly to advanced users with large projects.
Users often expect OCR software to be able to read less-than-perfect text, the way humans can read. Neither TextBridge nor any other OCR product quite does that yet. But TextBridge is impressive on good documents and can be trained and automated. You won’t find that anywhere else at this price.
Seiter, Charles. (September 1994). TextBridge 2.0. Macworld. (pg. 93).
Xerox Imaging Systems, Inc.
Applications
out of 10 (0 vote)
6 / 
2021-11-28 / 
3a1f4a3118e1b80b557d28728a95cdfc8b039055 / 
/ 
TextBridge_2.0.5.sit
Emulating this? It could probably run under: 
Basilisk II
