I'm working on scanning in a bunch of tax documents which can't be given to
me electronically. Some property descriptions, and lots of numbers. For the
most part, Tesseract is a godsend, saving me from retyping a lot. It seems
like the biggest part requiring manual correction is missing decimals. If I
look at the source PDF, they're clear as day, if admittedly on the small
side. They make it through the Ghostscript conversion to TIFF, and up to a
limit the resolution of the conversion doesn't matter (low resolution makes
everything worse, of course, but I stay away from that). When I tell
Tesseract to output the converted image (tessedit_write_images), I can
still see the little guys there, but Tesseract misses ~5%-10% of them.
Is there a good option to play with to bump up Tesseract's sensitivity with
respect to grabbing a decimal in, say, "76.50"? My own trial and error
isn't coming up with much. I thought I could use eng.user-patterns, but it
doesn't like wildcards in the first four characters, and that would seem to
be exactly what I'd want to do -- I feel like I'd want to use a Perl type
regular expression of ([\d]{1-3}\.[\d]{2}\-?) (1-3 decimals followed by a
decimal point and precisely two decimals, with an optional negative sign at
the end) but that wouldn't seem to be an option.
--
You received this message because you are subscribed to the Google Groups
"tesseract-ocr" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/tesseract-ocr.
To view this discussion on the web visit
https://groups.google.com/d/msgid/tesseract-ocr/8e7a25bb-5c9c-447d-b18f-a7b677beec14%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
