Optical character recognition S/W is getting pretty hot.
 It allows you to convert paper documents into ASCII form.
 Apparently even the low-cost OCR programs have 99% accuracy 
 on typeset of laser-printed pages.  And most have
 a recognition engine, and run their results through a dictionary.
 Recognition zones help the program divide and parse more
 complex page layouts.  They have Neural network technology which 
 learns and adapts so it can handle similar data in the future and 
 determine which conversion technique to use.  It's still
 kinda slow (as to be expected ), a 90MHz Pentium w/ 16Meg of
 ram takes about 1minute to scan and translate a page with a complex
 layout.  Some OCR S/W can preserve format and page layout.
 A few of the packages are Charactereyes Pro 2.51, Omnipage Pro 6.0,
 Textbridge professional edition 3.0 , Typereader professional 3.0XA,
 and Wordscan+ 4.0