Re^3: Extracting content text from PDFs

in reply to Re^2: Extracting content text from PDFs
in thread Extracting content text from PDFs

If there are limits to what PDFs work and what don't I have not run into them :)
I have not seen stringify send garbage to the output unless I tried to display a picture. For real text, it seemed to work just fine. I have no idea about those problems as it has worked for the uses to which I have put it.
sorry

Comment on Re^3: Extracting content text from PDFs
Re^4: Extracting content text from PDFs by almut (Canon) on Apr 16, 2008 at 03:32 UTC
If there are limits to what PDFs work and what don't I have not run into them :) You've just been lucky so far :) Some PS/PDF tools are using font subsetting/re-encoding techniques which (when done in a certain way) can make automatic text extraction very hard. (I've tried to explain the method in more detail in another thread.) To illustrate, here's a sample PDF which you can view with Adobe Reader, xpdf, Ghostscript, etc. without problems (you should see the standard "lorem ipsum" text). Any attempt to extract the textual content will likely fail, however, although the file is a perfectly valid PDF containing nothing but regular text content (no images, no encryption, no other tricks) with all characters being part of the ASCII set. Of course, I deliberately created the file in the above mentioned way (as "proof of concept"), but there are actually PDF creation tools out there which do produce such problematic PDFs.	[reply]

Comment on Re^3: Extracting content text from PDFs

Re^4: Extracting content text from PDFs
by almut (Canon) on Apr 16, 2008 at 03:32 UTC

If there are limits to what PDFs work and what don't I have not run into them :)

You've just been lucky so far :)

Some PS/PDF tools are using font subsetting/re-encoding techniques which (when done in a certain way) can make automatic text extraction very hard. (I've tried to explain the method in more detail in another thread.)

To illustrate, here's a sample PDF which you can view with Adobe Reader, xpdf, Ghostscript, etc. without problems (you should see the standard "lorem ipsum" text). Any attempt to extract the textual content will likely fail, however, although the file is a perfectly valid PDF containing nothing but regular text content (no images, no encryption, no other tricks) with all characters being part of the ASCII set.

Of course, I deliberately created the file in the above mentioned way (as "proof of concept"), but there are actually PDF creation tools out there which do produce such problematic PDFs.

[reply]