Publishers are thinking more about accessibility, but there is, understandably, a lot of confusion around how to do it. We attempted to break down some of those barriers in our recent webinar with Bill Kasdorf, What you need to know to get accessibility right.
As always, questions from attendees are illuminating for what the industry is thinking about and struggling to understand. In this post, we share the questions Bill received from publisher attendees (edited for anonymity, of course) and his response. We hope these provide insight for you as well!
More questions, or interested in learning more about accessible publishing best practices? Get in touch.
Q: Is this going to be focused exclusively on eBooks? Or also electronic publications that are web-based?
A: Excellent question! It’s a common misconception that this is just for books. In fact it’s for any kind of publication, actually whether web-based or not. I wrote a piece in Research Information recently about how now is the time for scholarly journal content to get accessible (if you’re interested), including the fact that the upcoming release of Atypon’s Literatum platform will generate EPUB 3s automatically for journal articles—a big step toward accessible journal articles (finally!).
Q: We find that our customers (higher education students) still prefer PDFs over EPUBs because then their pages will look exactly like the instructors’ pages. (Many don’t know about page end tags.) Do you see PDFs as inherently inaccessible because you can’t resize, change font, etc.? Or can PDFs be accessible too?
A: PDFs can be made more accessible (usually with a lot of hand tagging and other laborious remediation), but still nowhere close to what an accessible EPUB 3 offers, for the reasons you mentioned (reflowing, changing fonts, etc.), which is why the standard has moved to EPUB 3.
But you’re right, one of the issues the industry is struggling with right now is that students still ask for PDFs. That’s mainly because that’s what they’re familiar with and know how to deal with; plus many of the systems that deliver content to them (e.g., Kurzweil 3000) don’t yet handle EPUB well. (DAISY and a group of university DSS offices are working with Kurzweil on that, because Kurzweil recognizes that EPUB is better.)
Plus we’ve got to do a better job of making publishers and their suppliers know that EPUBs should have page break markers and the page-list <nav> — everybody assumes there are no page breaks in EPUBs (because usually there aren’t!) but as I mentioned in the webinar, Apex has been doing that routinely for years. One bright spot is that as newer generations of students come up, they are more inclined to EPUB; and I think you’ll see a sea change when Kurzweil gets its update out.
Q: Should the epub:type semantics be used in addition to ARIA roles?
A: That’s an open discussion in the industry right now. Some folks advise not to bother with epub:type anymore because few systems do anything with it. (The big exception is footnotes: iBooks and others use the epub:type attribute to trigger the popup footnote functionality.) Others point out that it’s useful to have those semantics just for content management purposes. (The @data attribute is sometimes suggested as an alternative, but there are debates about that too.)
Another issue is that epub:type is incompatible with pure HTML (EPUB uses XHTML), so when the Web Publication work that’s underway in the W3C issues its recommendation, epub:type is unlikely to survive that. But for now, it does no harm if it is useful in your workflow or your repository. I have advocated including it in most of the modeling I’ve done for clients the past few years, and am only now considering stepping back from it when appropriate.
Q: How does a Reading System mess up MathML in EPUBs?
A: The answer is all kinds of creative ways. Mainly it just botches the rendering—and as you can imagine, even if an equation is rendered 99+% correctly, the fact that something is wrong makes the equation wrong. And one equation being wrong makes a publisher mistrust the system enough to not provide MathML.
Another problem is that even when a publisher provides both MathML and an image of the equation in the EPUB, Reading Systems that _think_ they can render the MathML persist in doing so, even if incorrectly, rather than defaulting to the image for most (sighted) users. What we’re working on is figuring out how best to get them to default to the image in that case and leave the MathML to Assistive Technology. Benetech and DAISY are close to having a recommendation on how best to do that.
Q: When our typesetters are using MathType to typeset the equations, would it be problematic for them to supply the MathML? Most of our typesetters supply the equations as EPS files.
A: Not problematic at all, it should be completely straightforward. They can simply save them out of MathType as MathML as well as images and provide you with the MathML.
So please ask them to do that, even if they don’t put the MathML in the EPUBs for now!
Q: Should the alt text be alt=”” if the image is described in the caption? What is the best practice?
A: The key is whether it is _sufficiently well_ described in the caption. Remember that the user needs to know what the image is there to convey. So it depends entirely on the quality of the caption. Doing alt=”” is better than repeating the caption; but best is a better description, if the caption doesn’t say much.
Q: For “good enough” alt text, how about using semantic enrichment on the caption and in-text citation/mention of the image to provide appropriate keywords from a thesaurus? This is for a situation in which asking authors or third parties to write new custom alt text is impractical.
A: Very interesting idea and I believe this question was highlighted in the webinar, so thanks a lot for asking it! While this is not commonly done today, I think this is a promising route to pursue, now that AI and Natural Language Processing are getting so much more common and available.
However, keep in mind that just supplying keywords isn’t the same as a _description_. That’s why I mention AI and NLP rather than simply picking terms from a thesaurus. I think you’ll wind up finding that a combination of building the request or requirement for the image description into your MS management/image acquisition process, and then perhaps using these kinds of techniques as a suggestion mechanism when descriptions are not provided, might wind up working out. Especially in STM content, where images are technical, I really believe the authors are the best people to describe them.
Thanks again for the question, though, it’s a really interesting direction to pursue!
Q: Re: ARIA tagging: Can you do any of this in InDesign? I know, e.g., you can add alt text to images in InDesign. What else? I guess I’m looking for tips on minimizing the amount of post-export work that needs to be done.
A: I’m not aware of InDesign having a capability to insert ARIA tagging, but I reached out to a friend of mine at Adobe who is high up and technical to see if I could get a better answer. It appears that indeed InDesign doesn’t currently provide that capability. But at least maybe we’ve helped to get it on their radar!
Q: Are there resources or guidelines for finding government money or grants for updating workflows or legacy content?
A: I’m not aware of any, but you’ve got my wheels turning. I reached out to three of the projects funded by Mellon a couple of years ago to create open source platforms that will be freely available to scholarly publishers (Michigan’s Fulcrum, California’s Editoria, and Minnesota’s Manifest). All three of them are building accessibility features into their platforms. And I know this issue is a priority for Mellon. So there may be resources on the way to help.
Q: How do we include MathML in an EPUB just for accessibility?
A: Ah, the key to the answer is that little word “just.”
Of course you can include it normally, along with an image, but most publishers are reluctant to do that in EPUBs going out to the retail supply chain because too often the MathML gets rendered wrong.
If you are supplying an EPUB that is _just_ going to DSS offices, for example, then they’ll want the MathML for AT and will be able to ignore the image.
The best answer, though, is to wait a bit, because the folks at Benetech and DAISY are very close to having a best practice to address this question that will let you make just one EPUB, with images of equations for standard rendering, and with MathML for assistive technology that is ignored (in favor of the image) for normal rendering. I believe they’ll be publishing that answer in the next couple of months.
I hope this is helpful (though I know it’s not, yet, really!).
Q: Are there any written guidelines out there for writing alt text?
A: Yes, the best guidelines and tools are the Image Description resources provided by Benetech’s DIAGRAM Center which was one of the resources on my last slide. You will find that VERY helpful in many ways—guidelines, tools, examples, good stuff!
Q: How do you keep pirates from taking content when it is accessible?
A: The key is the party doing the distribution. for example, one of the most important aspects of Benetech’s Bookshare is that they rigorously qualify recipients as being eligible for the DRM-free files, and it works really well. That is an enormously important part of the process.
On college and university campuses, the DSS offices do the same thing. All of these parties are VERY aware that if they don’t handle the files properly they won’t get them from the publishers!
One other point: People often forget that it’s not usually the publisher that puts the DRM on an EPUB, it’s the retailer (Apple, Amazon, etc.). So the EPUB you send to the supply chain can still safely go to trusted recipients like Benetech or a college DSS office.
Q: In EPUB, we often reflow text from tables and figures into a nontabular format. University customers who have reviewed the files find that practice acceptable. What’s your opinion?
A: That is very helpful. While it obviously isn’t as good as giving that text in some sort of tagged or structured form, it still gives them something to work with. My guess is that the DSS offices on those campuses are taking your text and then doing that tagging. So the more you can keep some organization to the content, the easier you’ll make it for them, and in some cases your text file might be sufficient.
Q: Tables with multiple columns often display very poorly on devices with small screens. To overcome this, some publishers convert a range of tables to images – e.g. those with 3 or more columns. Even if these images of tables are accompanied by high-quality alt text and long descriptions, the accessibility is reduced. What are your thoughts on tables as images?
A: The heart of the answer gets at the difference between “access” and “accessibility” in the sense that we’re using the latter term in the webinar. The images are not accessible in that sense, and so high quality alt text and extended descriptions are very helpful. But the images are really for enabling sighted users to access tables on mobile devices, a different issue.
The best answer is both. So for example, if in your workflow you have tables tagged properly and can provide HTML tables, you can include that HTML as well as a fallback image. Even then, the high quality alt text and extended descriptions are good, because as anybody knows, oftentimes a reader doesn’t actually want to _read_ all the cells in a table, she just wants to know what information that table conveys.