Product Reviews

ScanSoft has released the first upgrade in two years of its OCR software package, TextBridge Pro, and in the process has made the leap from version 3.0 to version 8.0 to emphasise full compatibility with Mac OS 8 and to keep it in step with its recently released rival, OmniPage 8.0 (Reviews, Vol 14 No 5, p38).

Despite this numerical jump, changes to TextBridge Pro initially appear limited. By default, the program's main toolbar displays the same small but intuitive interface as previous versions. There are the same standard icons for selecting image source (scanner or image file), choosing preview options, deferring OCR (saving the document as an image file for batch processing later), and OCR training options (letting you interrupt the scanning recognition process to correct recognition decisions and improve the software's accuracy). Scanning remains simply a matter of inputting the original's quality and selecting configurable options for input and output layout as well as page orientation. These can all be chosen from a simple preferences panel in the main toolbar.

But what really matters to TextBridge users is whether the OCR engine is quicker or more accurate. Although scanning and recognition speeds appear only marginally better, TextBridge Pro 8.0 has measurably improved recognition accuracy, particularly when it comes to traditionally difficult originals such as faded faxes or smudged photocopies. TextBridge Pro recognised a simple text-based fax with near 100% accuracy, only stumbling on tiny, smudged print. OCR is a time-consuming process, so it's important that recognition is accurate, and the program managed to achieve high levels of recognition without much user intervention.

ADVERTISEMENT

The program shares its predecessors' frailties when it comes to recognising more unusual characters, such as pound signs. However, the built-in training facility, where you can correct common OCR errors as you move through the document, works very well: the number of errors that the OCR makes plummets after a few tutoring sessions. Some of the more esoteric zonal controls found in OmniPage are lacking here, but that's an omission most can comfortably live without.

We scanned a newspaper original, opting to export it directly to HTML. The results were respectable and the program correctly identified most graphics and text, but it still needed some tidying up - it would probably take less effort to export the file to plain text and then manipulate it in a Web authoring program.

Although recognition is not particularly fast, there are a number of time-saving utilities. For example, if you're scanning a large number of similar documents, you can defer recognition and save a zonal template, which means the same text and image recognition options will be applied to every scan in a batch. TextBridge Pro's AppleScripting options are equally impressive.

While AppleScript might be dismissed as too high end for most users, when you consider the time taken to scan and recognise complex documents, the ability to totally automate the process can be a lifesaver. A couple of scripts are included (there really should be more) with the program, and the documentation provides detailed listings of available AppleScript instructions.

Comparisons with its rival OmniPage are inevitable, particularly now both have settled at the same version number. While OmniPage has traditionally had the edge when it comes to OCR accuracy, the difference is now marginal. ScanSoft claims an accuracy improvement of 82% over previous versions, and although such claims are difficult to verify, TextBridge now copes superbly with deciphering smaller font sizes and italicised text.

TextBridge Pro 8.0 isn't foolproof, but at least it's got its priorities right. This upgrade is a significant improvement and, coupled with its low price and competitive upgrade options, it should be considered even if you already have less powerful OCR software bundled with your scanner.