I've made another attempt at this now. I'm a bit more hopeful about
this one, but still not sure.
In the new code, I write the text in a <p> element which is hidden, and
use aria-labelledby to say that both the <div> and the <canvas> are
labelled by that text.
I am quite hopeful that the text will be detected by a screen reader
now. You may hear it twice; if so, I have a couple of ideas how to fix
that.
I've also thought about what you said about the default text, and have
changed it slightly. Now the first plot in the page
https://dmurdoch.github.io/rgl/dev/articles/rgl.html
will have alt text "Rotatable plot with label unnamed-chunk-1".
Obviously customized text would be better, but I think this should be
better than nothing.
For authors using this, the R Markdown optional fig.alt setting almost
works. It's fine if you only have one plot, but if you have multiple
plots with different fig.alt settings rgl will only use the first one.
I think a change in knitr is needed so that rgl can work out which one
you really want.
The Shiny support is also a little deficient. It would make sense in an
interactive Shiny page to be able to change the alt text in response to
user actions, but that's not possible yet. If there are any Shiny
experts reading this, I could use some help with that.
Duncan Murdoch
On 17/03/2023 7:59 p.m., Jonathan Godfrey wrote:
Hello again Duncan,
I could see the two successes were <img> ... </img> creations and that the
others weren't.
I worry though that in searching for a solution, you need to end up with
something that is as simple for authors as what little effort is needed for
fig.alt; having looked at the HTML, I'd rather know what is happening at the
Rmd file stage and how anything created there gets passed on.
IMO: Your reflections on the content of the tags is bang on target. In a perfect world,
this is where authors' efforts should be, not on the technical "how do I get that
done" aspects.
I suspect that fixing the toolkit is necessary. A demonstration that a problem
can be solved is often a first step in creating access. Furthermore, this is
not a problem unique to your single document. Other authors will want to use
the same tools which you are demonstrating.
Alt tags are a perfect demonstration of access that does not impinge on
everyone who doesn't need that feature. Use of fig.cap (initially) and latterly
fig.alt to provide the access was done at the toolkit level, and it cannot get
any easier for authors. That maximises the chances that authors do then turn to
the content of the tags.
I'm happy to assist on projects that lead to positive improvements under the
hood that ultimately create access without authors having to become experts in
providing access.
We might take the testing off list, but keep the discussion of access on list.
Jonathan
-----Original Message-----
From: Duncan Murdoch <murdoch.dun...@gmail.com>
Sent: Saturday, 18 March 2023 12:20 pm
To: Jonathan Godfrey <a.j.godf...@massey.ac.nz>; R Package Development
<R-package-devel@r-project.org>
Subject: Re: [R-pkg-devel] Screen reader help request
Thanks very much for your response.
The two images where the tags showed up are regular figures, using lattice and
base graphics. There are also a lot of rgl figures, which have HTML code like
this in the source .html file:
<div id="rgl61108" style="width:480px;height:296.662546353523px;"
class="rglWebGL html-widget "></div>
(Looks like I should think about rounding the height value!)
All the interesting stuff (including the alt tags) is added within the div by
Javascript that runs after the page is loaded. The fact that it didn't show up
in your reader indicates to me that either it needs to be there before the
Javascript runs, or that I put the text in the wrong place. I'll try a few
variations on the code and hopefully find something that works.
Your comments about the content of the tags was also very helpful. I am
imagining that there are at least two groups of readers who might make use of
them:
- readers who would skip over the graphics completely, but who would be
helped by knowing what they missed.
- readers who can get information from the graphics but only with an
effort, so they would want to know where to apply that effort.
For the document I posted, there might not be very many people in either group,
but I'm hoping to make the package useful to others, who would be writing
documents with different audiences.
Duncan Murdoch
On 17/03/2023 5:18 p.m., Jonathan Godfrey wrote:
Hello Duncan,
I guess a few people might expect me to contribute to your request. First I'll assure
you that two graphics do have an alt tag which my screen reader is telling me about,
so in a sense that is success. I did not check with multiple screen readers or
different browsers because I would only find fault if I used inferior software; I
expect success with JAWS or NVDA to represent the other and likewise for browsers.
These two successes are the lattice and base graphics relating to the volcano data.
Others are not showing to my screen reader (JAWS) and browser (Chrome) on my OS
(Windows). I note that these are inserted using <img> though.
What should appear in an alt tag is an area of much debate though, and the attitude towards their creation differs widely. I believe that the context is the primary driver for what should be communicated. In your context, where you are obviously writing about the package (in general) and comparison of lattice v base graphics (where I do see the alt tags), it is the main text that should be communicating with your audience. You know what you are writing and why. If you do not write about the differences then you leave it up to chance that your (mostly visually-dependent) audience sees the differences that matter to you and your efforts as an author/developer. Complementing that information intended for the whole audience with a sensible alt tag on the graphics for the blind part of your audience, would be fundamentally different to what you might need to offer for the same graphics in another context, say an analysis feeding into a business report for example. In your context, the
blind person needs to know that the graphic does appear, and that the graphic
has a unique (enough) label to make it obvious which graphic appeared where. We
do not need a full description of every last detail of each graphic, although
in other contexts, that is exactly what we need.
Generating alt tag text automatically (no human interaction) is going to
deliver on the technical aspects, without the analytic outcomes, although our
recent efforts with the BrailleR package are trying to offer more and more
factual information that could help with the analysis of the data. Our main aim
is to assure the blind producer of a graphic that it does actually exist and
what (roughly) it shows; it would not be great text for an alt tag in most
circumstances. I've done some limited testing with AI solutions but haven't yet
been overwhelmed by positive experiences.
Anyone using an R markdown workflow can add alt text using the fig.alt option
in the header of their chunks. Without that, the fig.cap option is used, and if
neither is used, the screen reader user is victim to their personal settings.
(We choose whether to be told about unlabelled graphics which used to be
helpful for ignoring advertising.) I don't put unique alt tags in all of my
work because I just need to convey that the graphic is there, but I do make
sure there is a default bit of text that saves blind users with the wrong
settings from their decisions.
An author wishing to support blind consumers of their efforts, could spend lots
of time working out what to write, and never get it perfectly correct because
the desires of those consumers are not homogeneous. I advise people to use an
alt tag, but to make sure it does not duplicate what is already said in the
main text of their document. Authors already do this when thinking about the
right caption for their figures, but it is easy to see how varied that can be.
At least the target audience of captions is well known to the author given the
author is part of that population; in contrast, most authors are not blind and
may over or under-estimate the needs of the blind readers of their work. Before
doing everything an individual requests, please make sure that what is being
sought is in keeping with what you are hoping to communicate.
I despair at the retrospective creation of alt tags, especially if applied to
living documents. Creation of alt tags is simple for anyone using regular
markdown based insertion of their graphics because the ![]() construct embraces
them in the []. Authors using a LaTeX workflow must load an additional package
and add an optional argument to their \includegraphics{} commands. MS Word
users can label their graphics anytime but of late there is an increased use of
an automated alt tag creation for them (and messages in Outlook too for that
matter). I am loathed to fully describe the LaTeX pipeline unless the author is
also committed to producing their output in HTML as I have utter contempt for
the pdf (or post script) that is the default for most authors using any TeX
workflows.
Post hoc editing of html is perhaps the worst possible action I observe. This process has
separated the creator of the content from the creator of the access for blind people, and
is heavily reliant on the skill set of the accessibility "expert". Such editing
is common in university settings where the Disability Support Service is responsible for
the accessibility. If I was an undergraduate today, I'd be camping outside the doors of
my teaching staff to get the alt tags inserted. I spent plenty of time back then, sitting
in corridors waiting my turn, but so many of the access challenges of the dead-tree era
are thankfully behind us.
I'm curious though that a blind person is going to benefit much from your
package given so much of what I read relies on using a mouse, or having the
ability to imagine the 3d data projected into an inherently 2d
page/screen/image. In turn, that inaccessibility might heavily influence my
effort to provide comprehensive alt tags.
I'd be more than happy to continue this conversation, and to do so on list
because I think there are so many opportunities on offer today for people to
embrace; many of which will not create piles of extra work.
Jonathan
-----Original Message-----
From: R-package-devel <r-package-devel-boun...@r-project.org> On
Behalf Of Duncan Murdoch
Sent: Saturday, 18 March 2023 8:26 am
To: r-package-devel@r-project.org
Subject: [R-pkg-devel] Screen reader help request
I received a request a few days ago to add alt text to the plots in the HTML
vignettes for the rgl package.
That was easy to do, and I did it, but realized most of the graphics that rgl produces are
not in <img> format, they are WebGL canvases, and those don't support the same method
of adding alternate text. Online I've seen a variety of suggestions about how to do it for
<canvas> elements, and have implemented one of them.
Could I ask someone who is familiar with one or more screen readers to test
whether the method that I used works? A sample web page produced with the new
code is here:
https://apc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdmur
doch.github.io%2Frgl%2Fdev%2Farticles%2Frgl.html&data=05%7C01%7CA.J.Go
dfrey%40massey.ac.nz%7Cc0745ee90c414eebafa508db273e28e6%7C388728e1bbd0
437898dcf8682e644300%7C1%7C0%7C638146920163482319%7CUnknown%7CTWFpbGZs
b3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D
%7C3000%7C%7C%7C&sdata=tgPjXHerM93yrUrb963Tsa1MwsY7AKbFzaAFP8ZRYc8%3D&
reserved=0
The very first graphic is a plot of the iris data, and it should have alt text reading
"rglwidget in chunk unnamed-chunk-1".
That was default text generated without user input, users could put whatever
they wanted there instead.
Some related questions:
- Could you suggest better default text to use?
- Could you suggest good custom text for this particular graphic?
- The text is inserted when some Javascript has run to initialize the page,
so if the Javascript fails, there'll be no alt text. Is that a reasonable
limitation?
Duncan Murdoch
______________________________________________
R-package-devel@r-project.org mailing list
https://apc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fstat
.ethz.ch%2Fmailman%2Flistinfo%2Fr-package-devel&data=05%7C01%7CA.J.God
frey%40massey.ac.nz%7Cc0745ee90c414eebafa508db273e28e6%7C388728e1bbd04
37898dcf8682e644300%7C1%7C0%7C638146920163482319%7CUnknown%7CTWFpbGZsb
3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%
7C3000%7C%7C%7C&sdata=DTNqoAST5GrdK70Y%2B6X0gSOgn0dQQoXEN6cIR608YyE%3D
&reserved=0
______________________________________________
R-package-devel@r-project.org mailing list
https://apc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fstat
.ethz.ch%2Fmailman%2Flistinfo%2Fr-package-devel&data=05%7C01%7CA.J.God
frey%40massey.ac.nz%7Cc0745ee90c414eebafa508db273e28e6%7C388728e1bbd04
37898dcf8682e644300%7C1%7C0%7C638146920163482319%7CUnknown%7CTWFpbGZsb
3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%
7C3000%7C%7C%7C&sdata=DTNqoAST5GrdK70Y%2B6X0gSOgn0dQQoXEN6cIR608YyE%3D
&reserved=0
______________________________________________
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
