Making books readable on computer proves trying task

This article was written in December 2004 and gives an overview of the problem with Optical Character Recognition (OCR). If you haven't thought much recently about Optical Character Recognition, you should. Although it has gotten better, it is still not perfect. Dean Tang, CEO of OCR software-maker ABBYY USA Software House, says that on perfect text, OCR can top 99% accuracy. That means out of 1000 words, 10 will be incorrect. No bad, but what do you have to do to find and correct those 10 mistakes? And if you don't correct them, how will text searching be affected?