posted 14 years ago
It wouldn't be *that* difficult, really. Relative to something really, really difficult, anyway.
The OCR part is largely a solved problem. And some of the commercial OCR bits will let you select chunks and drop them into tables--I assume they just identify regions (lines, which is trivial, and columns, which is probably just the same thing with some wiggle room for the unjustified text). If you can get that far, you can turn it into CSV.