[accessibility] Testing PDF tagging with a pdf generated using pdflatex

[Ross Moore]

Hello Tammy,

On 16/03/2018, at 7:56, "Stitz,Tammy A" <tstitz at uakron.edu<mailto:tstitz at uakron.edu>> wrote:

Hello Everyone:

I know that some of you wanted to know what happened when I tried to make a tagged PDF from a LaTeX generated PDF, so I am sending this email to the list. I tried a few packages when using pdflatex, but none seemed to make a structured, tagged PDF. The only software that I know that can add tags to a PDF is Adobe Acrobat.
Yes, it does a reasonable job, but needs a lot of touching up afterwards.
It is this manual post-processing that makes it unsuitable (yet) for an automated production system.
But it is important that people like yourself try it out, to appreciate both the strengths and weaknesses of this approach — that is, adding tags after all the typesetting has been done.


If you have used any other software, please let me know.
What is really needed is for LaTeX processing to add tagging, according to the structure inherent in:
 1. the author's LaTeX source
 2. the chosen document-class
 3. any loaded packages

This is what I am working on, and have achieved with several significant examples, chock full of mathematical content. These will be part of my TUG presentation in Rio, later this year.

They are not yet available online.
I'll post again when they are.


When Acrobat generated the tags, some strange things happened. (1) Some brackets and other large symbols like summations disappeared when I had Acrobat assign tags
This happens with TeX's extension fonts, when there is no /ToUnicode resource.
It helps to load  glyphtounicode.tex   to get mappings based upon the glyph name.
This is just one of the things done when the  pdfx  package is loaded.

The first step in developing methods to produce tagged PDF compliant with published ISO standards, is to first satisfy all requirements for PDF/A compliance. There is significant overlap.
Ultimately you want to satisfy both  PDF/A-2a and PDF/UA-1.
The  pdfx  package is vital to be able to achieve this.

. I was thinking it might have to do with when a font was compressed. I adjusted what I used for commands (e.g. use bmatrix instead of \left\right). I was unable to find a pattern, but I thought it was worth mentioning.

Many tags and much of the reading order needs corrected after Acrobat creates the tags.
Precisely. It's too much work, and cannot result in anything close to what a fully-automated approach from LaTeX can achieve.


If you have mostly text and simple equations, the screen reader JAWS reads it well. I used my thesis as a test case and it had many matrix equations, which are read awful by JAWS. The only solution that I could find was to add alternative text to each equation, which was very time consuming
The big question here is  ‘What should that alternative text be?'
Just reading the characters separately is not sufficient.
How about the LaTeX source, complete with backslashes ?
Some people say that this is desirable.
What do you think?

. I wanted to find a simple solution that students would use when they submitted their theses and dissertations. I know this process can't take much time or effort. I think of one of my students, who was an excellent student, that said, "Could you read all the documentation for datatool and tell me the highlights"? Uh, no.
Maybe that will be possible soon.  3-5 years hence?


Here is an example of how JAWS reads without alt text and a matrix equation (video)
https://youtu.be/uN0RhcQw8RI
Here is an example of how JAWS reads with alt text and a matrix equation (video)
https://youtu.be/M4w_8yP88II
My tagged thesis posted on our institutional repository (PDF).
https://works.bepress.com/tammy_stitz/6/

I'll have a look at this.
Thanks for the link.


All the best,

   Ross

Best Regards,
Tammy Stitz
Associate Professor, Bibliography
Applied Sciences Librarian
Email: tstitz at uakron.edu<mailto:tstitz at uakron.edu>
Office: 330-972-6192
Science & Technology Library
The University of Akron
Akron, Ohio 44325-3907
http://www.uakron.edu/libraries
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://tug.org/pipermail/accessibility/attachments/20180316/c9a61727/attachment-0001.html>