WebAIM - Web Accessibility In Mind

E-mail List Archives

Re: Untagged PDF doc with table structure

for

From: Chagnon | PubCom
Date: Feb 18, 2015 12:41PM


Lynn, I too have a strong programming background in HTML, as well as SGML,
XML, and many other markup languages. So tags plus reading order create the
document's structure in my mind! In theory, I don't believe a PDF can have
any structure, good or bad, without tags. All PDFs have a page architecture,
but that's not the same thing as structure.

Lynn asked: " if so how would I recognise it if I were to examine the
document's building blocks "

You have to examine it from several viewpoints in Acrobat Pro. I teach my
students this method:
1. Run Acrobat's accessibility checker. This looks at only about 20% of the
document's features, so don't depend on it for a full check.

2. Run down the tag tree, top-to-bottom. I call this the tag reading order.
For sighted users, they can arrow down from tag to tag and also see on the
page which item is highlighted for each tag. They'll see very quickly that
the figures weren't read at the correct place in the tag tree, or that the
second half of body text was read first, then the heading 1, then the
remaining body text.

For screen reader users, this is what your software is using. But it's more
difficult to tell if the document is correct. Were you able to hear and
figure out what was read? Did it make sense (not the content itself, but the
order in which you heard it)? Screen readers also can't tell sometimes if
it's tagged correctly. Example: Adobe InDesign has a tragic flaw. When a
sidebar (boxed text that's secondary to the main story) is exported to PDF,
the conversion isn't correct. All of the text is jumbled together;
paragraphs are lost, including any headings, bulleted lists, tables,
figures, etc. So a screen reader just hears the text run-on blah blah blah,
but never knows if he's reading one paragraph, multiple paragraphs,
headings, or any other parts of a document. My screen reader testers often
miss these problems; they just can't tell if they've missing something or if
it's incorrect.

3. Run down the "real" reading order. This is the Order panel in Acrobat.
Often overlooked by many in accessible documentation, this is the original
reading order that's still used by many assistive technologies, including
braille printers and keyboards. I've never had any of my screen reader
testers review this because their software has a hard time voicing it in a
way that makes sense to them. But they can see this reading order another
way; View / Zoom / Reflow. This utility rejiggers the visual layout on the
screen to mimic the real reading order. Columns are removed, everything is
sequential and linear, top to bottom. So if the first item read by a screen
reader happens to be the photo caption, not heading 1, then you have a
reading order problem.

4. After that, the usual review of tags, tables, alt-text, etc. takes place.

--Bevi Chagnon

-----Original Message-----
From: <EMAIL REMOVED>
[mailto: <EMAIL REMOVED> ] On Behalf Of Lynn Holdsworth
Sent: Wednesday, February 18, 2015 1:46 PM
To: WebAIM Discussion List
Subject: Re: [WebAIM] Untagged PDF doc with table structure

Hi Bevi,

Thanks for taking the time to write such a comprehensive response.

From creating HTML pages for about half a lifetime, I'd define tags and
structure pretty much the way you do.

But I inferred from this thread, and from talking with someone who knows a
lot more about PDF than I do, that it's possible to have structure without
tags in a PDF document. Is this correct, and if so how would I recognise it
if I were to examine the document's building blocks?

Best, Lynn

On 18/02/2015, Chagnon | PubCom < <EMAIL REMOVED> > wrote:
> Lynn wrote: " in PDF docs, what's the difference between tags and
> structure?
> "
>
> This is one of the toughest concepts we teachers have to explain! I'd
> love to hear how others describe it. Here's my take:
>
> Tags are labels. Code labels, specifically, that are read by Assistive
> Technologies and are not usually visible to sighted users unless they
> have Acrobat Pro. They let AT users know what's a heading 2, a list of
> bullets, tables, and other parts of the documents. Tags also do a lot
> of work for us, such as assisting us in creating bookmarks and tables
> of contents, creating navigation systems, and holding the Alt-text on
> graphics (Alt-Text is an attribute on the figure tag and doesn't stand
> alone on its own).
>
> Structure is the sequence of how the document's pieces will be read,
> or in other words, the sequence in which the tagged items are read.
> Call it reading order or tag reading order. The structure of some
> documents can also have nesting qualities, such as all the pieces of a
> chapter, and all the chapters in a book.
>
> An example: If Heading 1 designates a chapter title, then all the
> paragraph, bullets, tables, and heading 2 items within that chapter
> will be nested inside the main heading 1 tag. This allows AT software
> to figure out, hopefully, what goes with what; that all the tags
> nested within Heading 1 is a chapter.
>
> Structure is created when you have tags (the right tag labels) and a
> reading order (a logical reading order). It is possible that a tagged
> and structured document might not be fully accessible because the tags
> aren't accurate enough or the reading order is out of whack.
>
> Example number 1: In older versions of MS Word, figures would be
> placed in very odd places of the reading order when it was exported to
> a PDF. If paragraph 1 stated "see figure 5", figure 5 itself might end
> up at the very end of the reading order, not near paragraph 1 where it
> was referenced. A sighted person sees figure 5 next to the paragraph,
> but a screen reader user doesn't hear it voiced until the last page,
> and maybe that's page 360 of a long government document. So the
> document is tagged and structured, but it's a faulty structure because
> the reading order is incorrect.
>
> Example number 2: Graphic designers who use desktop publishing
> programs like Adobe InDesign and QuarkXpress create very complex
> visual layouts.
> Visually,
> things aren't designed in a traditional top down left right pattern
> but instead could be scattered all over the physical page. Here's an
> example of a 2-page magazine spread:
> http://fc02.deviantart.net/fs71/i/2010/082/e/c/Magazine_Layout_Design_
> 1_by_B reakTheRecords.jpg (This is just a random sample I pulled up
> on the Internet, so it is only a graphic of a 2-page spread, no live
> text or
> Alt-text.)
>
> Note that article title (or heading 1) appears on page 2, and the body
> text of the story starts on page 1. Backwards! And then there are 2
> quotes at the top of page 1, so obviously the designer wants us to
> read those at the beginning of the story, also. And here's a similar
> example:
> https://m1.behance.net/rendition/modules/12455236/disp/322ee0c042b2949
> 607393
> d8b1f24ad96.jpg
>
> Whew! Getting a tagged, logical reading order from this type of
> publication isn't easy!
>
> Summary:
> Structure equals tagged content placed in a logical reading order.
>
> Well, that's my attempt. Would love to hear how others describe the
> concepts.
>
> --Bevi Chagnon
>
> -----Original Message-----
> From: <EMAIL REMOVED>
> [mailto: <EMAIL REMOVED> ] On Behalf Of Lynn
> Holdsworth
> Sent: Wednesday, February 18, 2015 12:11 PM
> To: WebAIM Discussion List
> Subject: Re: [WebAIM] Untagged PDF doc with table structure
>
> Thanks so much everyone for weighing in - I've found this a very
> useful thread indeed.
>
> One more question: in PDF docs, what's the difference between tags and
> structure? Ryan mentioned that the doc may include structure but not
> be tagged, and I don't understand the difference.
>
> And thanks Duff for the LinkedIn group suggestions. I'll join at least
> the first one.
>
> Really hoping that Adobe is working on ironing out the accessibility
> glitches in the DownLoad Assistant, as I'd appreciate the chance to
> learn about and use what seems like a great bunch of accessibility
> features in Acrobat.
>
> Best, Lynn
>
> On 18/02/2015, Andrew Kirkpatrick < <EMAIL REMOVED> > wrote:
>> Bim,
>> I was talking about both Acrobat and Reader in my reply, sorry if
>> that wasn't clear. It is the same process for both.
>> AWK
>>
>> -----Original Message-----
>> From: <EMAIL REMOVED>
>> [mailto: <EMAIL REMOVED> ] On Behalf Of Bim Egan
>> Sent: Wednesday, February 18, 2015 7:13 AM
>> To: 'WebAIM Discussion List'
>> Subject: Re: [WebAIM] Untagged PDF doc with table structure
>>
>> Lynn didn't seem to be talking about using Acrobat though. She
>> described the experience of many screen reader users in finding a
>> table in an untagged
>> PDF when opened in Reader, and she asked why this could happen. Her
>> message said that the Acrobat installation wasn't accessible.
>>
>> Bim
>>
>> -----Original Message-----
>> From: <EMAIL REMOVED>
>> [mailto: <EMAIL REMOVED> ] On Behalf Of Andrew
>> Kirkpatrick
>> Sent: 18 February 2015 14:36
>> To: WebAIM Discussion List
>> Subject: Re: [WebAIM] Untagged PDF doc with table structure
>>
>> Jon is correct. When Acrobat opens an untagged document and there is
>> a client that is using the accessibility API data running, Acrobat
>> (or
>> Reader) will add tags to the document. The result is the same as if
>> an author used the "add tags" feature in Acrobat. You get Acrobat's
>> best interpretation of what the tags should be. That will sometimes
>> result in headings, well-formed tables, lists, and other structures.
>> Authors who use this feature in Acrobat know that you generally need
>> to
> fix some of the tags.
>>
>>
>>
>> The result is that the document is tagged temporarily and assistive
>> technologies recognize and use the information.
>>
>>
>>
>> The dialogs that you see when opening PDF documents give you some
>> information about what is going on. To understand better, here's my
>> explanation.
>>
>>
>>
>> In acrobat or Reader preferences there is a "Reading" category.
>> There is a checkbox that is labeled "Confirm before tagging
>> documents". If this is checked, then every time that Reader intends
>> to tag an untagged document the "Reading an untagged document with
>> assistive technology" dialog pops up and the user needs to confirm
>> that this is what they'd like to do. If the user selects cancel then
>> the document won't be tagged and the reading experience will be
>> essentially
> non-existent.
>>
>>
>>
>> If you elect to allow the tagging, there are other options as
>> mentioned in one of the replies. I recommend using the "infer reading
>> order from document" option.
>>
>>
>>
>> There are other settings related to large documents and auto-tagging.
>> Autotagging takes time, so if you open a very dense 600 page manual
>> you may find that Reader takes a long time to do the tagging. It
>> can, and we are always looking to improve the efficiency of this process.
>> The option for the user is to indicate whether the autotagging should
>> occur only on visible pages, on all pages in the document, or on all
>> pages except if the document is "large". The user gets to define
>> what large means - a user might find that their system is slow at
>> this so sets the limit at 25 pages, or might set it higher if their
>> system handles this process quickly. The down side of only tagging a
>> few pages at a time is that if there are recognized structures on
>> pages that haven't been tagged yet (e.g. a heading on page 51) the
>> user can't use screen reader heading navigation to jump to it because
>> the tags
> don't exist until the page is in view in the reader.
>>
>>
>>
>> Hope this helps,
>>
>> AWK
>>
>>
>>
>> -----Original Message-----
>> From: <EMAIL REMOVED>
>> [mailto: <EMAIL REMOVED> ] On Behalf Of Lynn
>> Holdsworth
>> Sent: Wednesday, February 18, 2015 4:36 AM
>> To: WebAIM Discussion List
>> Subject: [WebAIM] Untagged PDF doc with table structure
>>
>>
>>
>> Hi all,
>>
>>
>>
>> Apologies if PDF accessibility is off topic. If so is there a list
>> that covers this?
>>
>>
>>
>> But if not ...
>>
>>
>>
>> I open a PDF document, and Adobe Reader alerts me that it's untagged.
>>
>>
>>
>> So I begin to peruse it using JAWS, and come across a table whose
>> structure is robust enough for me to move around it using the JAWS
>> table
> keystrokes.
>>
>>
>>
>> Does this mean there *are* tags in the document after all? Or has
>> Adobe Reader used heuristics to add tags to improve the doc's
>> accessibility, since my settings flag up that I'm using a screenreader?
>>
>>
>>
>> I tried to download a trial version of Acrobat Pro so as to examine
>> the document structure, but the download assistant seems inaccessible.
>>
>>
>>
>> Thanks as always, Lynn
>>
>> >>
>> >> list messages to
>> <EMAIL REMOVED> <mailto: <EMAIL REMOVED> >
>> >> >> list messages to <EMAIL REMOVED>
>>
>> >> >> list messages to <EMAIL REMOVED>
>> >> >> list messages to <EMAIL REMOVED>
>>
> > > list messages to <EMAIL REMOVED>
>
> > > list messages to <EMAIL REMOVED>
>
messages to <EMAIL REMOVED>