The latest version of ReadIris Pro is a major improvement on its predecessor. It raises the number of languages supported to an amazing 117 – from Afaan Oromo to Zulu – saves recognised text in 14 text formats and six image formats and adds recognition for barcodes. Most important of all, it claims an increase in recognition accuracy of about 30 percent over Version 9.0. Most readers that have bought an HP scanner will no doubt already be familiar with the 'lite' version of ReadIris that comes bundled with most ScanJets.
A Corporate Edition is also available, at a premium of around about £170. As you'd guess it has a laundry list of additional features largely suited to the mass OCR-processing of documents, such as support for high-speed and duplex scanners and batch-OCR processing, plus niceties such as business-card and mixed-language recognition, and support for vertical market lexicons. The program comes complete with that modern rarity, a reasonably informative printed User's Guide manual, which to our jaded ways is preferable to a PDF file.
The program's previously plain interface has had a makeover and now sports a glossy XP-style interface. It still has only two panes: one for processing and another for the image. So, it remains easy to use. You won't find an editing pane such as those in OmniPage Pro or FineReader, because there's no built-in editor. Once you have recognised a set of pages, the program saves the file and opens it in the appropriate program. If the program allows editing, you can then look for and fix errors. Some may be happy with this approach but we preferred doing corrections before saving to a file.
The proof of the OCR pudding is the accuracy of the recognised text it delivers and here ReadIris is pretty much a match for its rivals – put a decent source document before its gimlet OCR eye and ReadIris will recognise most characters, very often getting 100 percent right. But these days, that's not saying much – most OCR packages can manage this with one virtual hand behind their back.
ReadIris Pro 10.0's performance on skewed source material, which, lets be honest, is a difficult chore for OCR packages, was distinctly weak and had no trouble in generating plenty of gobbledegook. By contrast, both OmniPage Pro 15.0 and FineReader 8.0 coped remarkably well, producing a very low order of errors indeed.
One of the more intriguing features of ReadIris Pro 10.0 is its ability to read handwriting or, to be more specific, hand printing. The ability to turn hand written words in to text is potentially very useful, but don't hold your breath – ReadIris Pro has some way to go before this feature works as advertised. To say it's picky about the way you write is an understatement – you have to use separate block capitals only. To help you, a printed example of specimen capitals is provided. We popped this under the scanner thinking that it ought to do well with the supplied example, but we were wrong – as you can see from the screenshot, pure gobbledegook was the result. So we didn't bother to try it with the usual doctor's prescription-style scrawl!
All in all, this was very disappointing. After all, Pocket PC PDAs – with their limited processing power – have been able to make a decent fist of recognising proper (cursive) handwriting for some time now, albeit with the advantage of being able to ‘watch' letters being written.