From: "Geo." 

It's less reliable if it's not scanning words from a paragraph of text
where it can use sentence structure and a dictionary to correct itself.
Give it a list of part numbers consisting of nothing but letters that make
no valid words and it won't do any better than it does with alphanumerics.

Geo.

"Ellen K." 
wrote in message news:q17r30dgev3lhunp2f3jp05ns2vb3n5l4r{at}4ax.com...
> In general, OCR is less reliable for numbers than for letters...
> depending on the fidelity of the scanned image I wouldn't be surprised
> if it mixes up 5 and 6 or 6 and 8 or 8 and 9.

--- BBBS/NT v4.01 Flag-5