Previous | Next | (P-PDF) PDF Accessibility
Topic: RE: OCR compatibility with tag tree/searching (Via Email)
Conf: (P-PDF) PDF Accessibility, Msg: 134329
From: Duff_Johnson
Date: 6/11/2005 02:07 AM
> I'm trying to create accessible, searchable PDFs from paper texts.
>
> I scan them as 300 dpi bitonal TIFFs and then perform OCR
> using OmniPage Pro 11. Then I save back as PDF with image on
> text. Retaining the image is important for these archival documents.
>
> The first problem I'm having is with the tag tree. In
> Acrobat 6.0 Professional, I use the Add Tags to Document
> option in the Accessibility portion of the Advanced menu.
> This invariably creates a Figure tag for the page image and a
> Table tag for the page of text. Each line (or portion there
> of) of text is given a Paragraph tag. Also, the graphic
> zones that I've created in OmniPage are not recognized as
> Figures in Acrobat. This, to me, makes the tag tree
> relatively useless. But, the Web Accessibility people at my
> institution teach "an accessible PDF is a tagged PDF."
They are correct!
First of all, the fact that you are saving your OmniPage file to
PDF/Searchable Image (Image on Text) is the factor causing the
image-zones you defined in OmniPage to disappear. Creating image-zones
is only meaningful for PDF/Formatted Text and Graphics (PDF/Normal)
output, which won't work for you on account of your need to retain the
entire page-image.
I would remove the page-image from the structure tree, then retag the
text so it's not in a table (unless it should be, of course).
> Do you have any experience with this or know of any solution?
> Do other OCR programs work better with Acrobat?
Well... Acrobat Capture 3.05 has old OCR, but it has smoother
integration with the tagging process, although the process is not at all
refined, and has some serious gaps.
The thing to focus on is getting the best-possible OCR results (not
software), then address the tagging in Acrobat. If accessibility is a
priority for your scanned document, then correcting the OCR to (a) clean
out "debris" from graphics and (b) ensure all the words will read
correctly is a must.
Having created a high-quality OCRed document (and this is NOT an
all-automation process), you can then tag them in Acrobat (Acrobat 7.0
Professional is better for this than 6.0 Pro).
> Also, some of my documents are math papers with formulas.
> The powers that be want the formulas to be expressed in
> natural language for screen readers as well as searchable by
> LaTex. Any suggestions on where to hide all of this text?
To do that right, you'd need to drop the PDF/Searchable Image model,
because it simply does not allow you multiple image zones on the page.
However, since I'm sure I can't talk you out of Searchable Image PDF,
then I'm afraid you are left with "hiding" the text as "Actual Text" for
some convenient text-element (such as a period). This is ugly, but it
works. However, I would really suggest considering Formatted Text and
Graphics for these pages.
> My
> current (but untested) solution is to put the natural
> language in the OCR layer and the LaTex in Bookmarks,
> although this seems less than ideal, and I'm not sure how it
> will work across platforms.
You mean... you are adding LaTex code as bookmark text?!? Could you
explain this a bit further?
Duff Johnson
Document Solutions, Inc.
http://www.document-solutions.com
