OmniPage Pro 9

The point of OCR is to end up with as accurate a reproduction of the original document as possible. To do this, the software must have the ability accurately to recognise the words in the document and be able to reproduce the layout of the page. Caere's latest version of its industry-leading OmniPage Pro improves on both aspects, but particularly on its ability to recognise and convert more complex objects like tables, colour graphics and spreadsheets.

Over its last few incarnations, OmniPage has featured greatly improved character recognition and a better user interface. It now claims to have more than 99 per cent accuracy on good-quality originals using standard typefaces. You could achieve this if you just wanted to scan and electronically store simple documents, such as letters and single column reports. However, the accuracy lessens with more complex documents, while decorative fonts prove particularly difficult for OmniPage. Lower quality originals such as faxes also reduce the accuracy, although OmniPage makes a surprisingly good job of recognising poor and even badly skewed pages. Of course, even at a lower accuracy, using OmniPage is still faster than retyping.

Installation requires at least 50Mb of free hard disk space, but you should aim for double that. An MMX-enhanced processor will perform much better than a standard Pentium and, ideally, you need a clock speed of at least 166MHz. OmniPage Pro recognises all common scanners and includes a standard TWAIN driver for those not on the list. It works best with SCSI scanners, particularly multi-feed models that OmniPage can use to scan several pages at once. Version 9 now includes support for multifunction devices such as the HP OfficeJet series.

The OCR process in version 9 is very similar to its predecessor. You can convert documents by using the OmniPage interface or put them straight into your word processor using the Acquire Text option, which is added to the file menu during installation. The latter option is easiest for simple, individual sheets, but the quickest way to scan in a series of pages intended to be part of one document is to launch OmniPage itself.

The most obvious improvement to the Auto-OCR Wizard in version 9 is an opening screen that asks whether you want to scan in black and white, greyscale or colour. OmniPage Pro 9 is the first OCR package to support colour graphics, which means the laborious process of OCRing the text and then manually scanning and inserting colour graphics should be a thing of the past. The downside of the colour option is that it takes about 20 times longer than scanning in black and white. On pages with one or two small colour graphics, it would probably be quicker to import the text through OmniPage and scan in the images separately. Clearly, the faster your scanner, the less this is an issue.

Once you've indicated how many colours to scan for, you can choose the basic layout of the document or let OmniPage do it for you. Version 9 recognises spreadsheets so they can be edited in Excel rather than as tables in Word. Next, you choose a language from the list of those you chose on installation. You also choose whether OmniPage should look out for tables as it can now output tabular information as Word table objects rather than by using tabs. Finally, you indicate to what extent OmniPage should represent the layout of the document. The more accurate the representation, the less editable the text will be.

Once you've completed the Auto-OCR Wizard, the process of scanning begins. This needn't take long; using my very average parallel port scanner it took just 30 seconds or so to scan in a black and white page.