Copy link to clipboard
Copied
Hi,
I have an odd issue with some PDFs that I have to extract data from, and I'm hoping for some support to unpick the problem.
The files appear to have text in them that is hidden in some way other than what might be considered normal. If the attached PDF is viewed in Acrobat, it appears ok but when I run a parser over it to extract the data from partular parts of the doc (using docparser.com, but I'm confident the issue lies in the PDF, not the parser) there is additional text appearing in parts of the doc.
An example problem area is the small table on the right hand side of page 2 titled 'Cott' with rows of 'Odd', 'Medium' and 'Heavy'. I'm extracting that particular area - just the table data, not the vertical title - with the parser, and there is text appearing in it that cannot be seen when viewing the doc in Acrobat.
I've tried the following in Acrobat:
Removing form fields
Running the 'Remove hidden text' operation.
Neither of these operations are finding that extra text and I'm out of ideas. Hoping that somebody can assist with removing anything that can't be seen without flattening the document to an image (resulting in the need to do OCR after, which is not an option in this case).
Copy link to clipboard
Copied
The problem is in fact not hidden text. None of the extra text is hidden. Rather it is on the page but in a different place. Take a look at this pic. (No pic in email replies).
Things extracted from the "Cott" box include parts of vertical captions, like "pe bre" which is made up from "type"..."break" just below the box. So this is one to give back to the text extraction people - or seek a different tool.
Copy link to clipboard
Copied
" I'm extracting that particular area - just the table data, not the vertical title - with the parser, and there is text appearing in it that cannot be seen when viewing the doc in Acrobat."
What text does you get there?
Copy link to clipboard
Copied
Copy link to clipboard
Copied
Apologies, ignore previous reply. Wrong screenshot attached. The text displayed from the document that's attached to this thread is as follows:
Copy link to clipboard
Copied
To analyse this - since Acrobat doesn't extract any extra text - we'd need the exact text you extract, and the exact options you used in your non-Adobe extractor (for example, do you extract by coordinate area)?
Copy link to clipboard
Copied
The parser requires a bounding box to be drawn on the file and extracts the text that it finds there. This is exatly what is extracted from the document attached to this thread:
Copy link to clipboard
Copied
The problem is in fact not hidden text. None of the extra text is hidden. Rather it is on the page but in a different place. Take a look at this pic. (No pic in email replies).
Things extracted from the "Cott" box include parts of vertical captions, like "pe bre" which is made up from "type"..."break" just below the box. So this is one to give back to the text extraction people - or seek a different tool.
Copy link to clipboard
Copied
That makes a lot of sense. Thank you! the bounding box of the selection area does not cover the vertical parts, but I assume that it may be something to do with the way that it's rotated when being read that's causing it to invisibly (to the eye) fall into the bounding box.
I'll go back to the parser team.