iText (to my limited knowledge) doesn't have a "Word Finder" like Acrobat does.
The problems with extracting text from some PDF found in the wild are as
follows:
1) The text could be in just about any encoding, including one created
on the spot for that particular piece of text. Determining this encoding is
possible, but requires some effort.
2) The text might be raw glyph indexes ("the Nth character in this
font" with no ordering guarantees of any kind) ... you'd have to crack open the
font file and hunt through it's character mapping tables to determine the right
character. I have yet to see a word finder handle this case.
3) They just might be raw-est drawing commands. Curve-to's and
line-to's. OCR is your only recourse at that point. I have yet to see anyone
handle this case.
4) Text doesn't have to appear in a contiguous block. There can be
kerning information between letters (information to adjust the spacing between
letters so things like 'ij' or 'll' look better). Each letter can be drawn
individually... heck, it's perfectly legal to draw all the characters in
alphabetical order rather than by location. Inefficient (lots of moving the
current drawing point around), but valid. The end of a run of characters can
appear at any time... cutting words in half.
"Word Finders" like the one found in Acrobat/Reader have to figure out where
all the letters are on a page, what those letters are, and then build words out
of them based on their position (letters sharing a base line with only X
distance between them are part of the same word... that sort of thing).
But that's the worst case scenario... Some random PDF build by some random
application... it has to work with anything that's legal PDF.
You're not in that scenario.
The PDFs produced by the IRS will be from a limited number of applications...
possibly even "1". Examining the raw output will show you short cuts that,
while handy for your particular case, would be Really Bad in the general case.
Somthing in the GhostScript family may have a word finder. Poking around
revealed that GSview claims to be able to search for text (which requires
knowing how to find words). http://www.cs.wisc.edu/~ghost/gsview/. GSview is
released under the GPL.
You may even be so lucky as to have PDF Structure in your PDFs that
specifically calls out the text of each paragraph... for things like
text-to-speach software. The gub'ment is big on accessibility-enabled PDFs.
At that point, you don't really need to worry about what's drawn on the page at
all, you can just poke around in the Structure tree (still work, but not so
daunting).
--Mark Storer
Senior Software Engineer
Cardiff Software
#include <disclaimer>
typedef std::Disclaimer<Cardiff> DisCard;
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf
> Of Richard
> Braman
> Sent: Tuesday, February 14, 2006 10:35 AM
> To: itext-questions@lists.sourceforge.net
> Subject: [iText-questions] Reading and Extracting Text from PDF
>
>
> I have a open source project that is attempting to structure IRS
> produced documents such as publications and instructions and parse out
> data that is critical to building tax software.
> An example of such a file is http://www.irs.gov/pub/irs-pdf/p1346.pdf.
> This file contains e-file record layouts, which start on page
> 398. They
> used to publish this as text which made parsing relatively
> easy, but now
> it comes in PDF only, and the project needs to be able to
> have good open
> source parsing technology. Is Itext the right tool for this job? I
> have seen it do good work on parsing the metadata contained in IRS
> fill-in forms.
>
>
> Richard Braman
> mailto:[EMAIL PROTECTED]
> 561.748.4002 (voice)
>
> http://www.taxcodesoftware.org
> Free Open Source Tax Software
>
>
>
> -------------------------------------------------------
> This SF.net email is sponsored by: Splunk Inc. Do you grep
> through log files
> for problems? Stop! Download the new AJAX search engine that makes
> searching your log files as easy as surfing the web.
> DOWNLOAD SPLUNK!
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&;
dat=121642
_______________________________________________
iText-questions mailing list
iText-questions@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/itext-questions
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid3432&bid#0486&dat1642
_______________________________________________
iText-questions mailing list
iText-questions@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/itext-questions
