There's a new sheriff in town …

[elibrarian]

Amongst the features of epub 3 is a possibility to insert accessibility tags to make the files more accessible for reading tools for handicapped. The possibility in 3.0 gets somewhat enforced in epub 3.1. No-one here at MobileRead seems to care much, and most of the ebooks I find "in the wild", that calls themselves epub 3 doesn't either. The only epub-creator tool giving the possibility to insert such tags (at least some of them) I have found are Jutoh.

But now there seems to be a new sheriff in town, so to speak, since the DAISY consortium are making an "accessibility checker" on the lines of epubchecker (https://daisy.github.io/ace/), that are now in the release candidate stadium. I guess that the ebook-distributors will demand new epubs to pass this as well as epubcheck in some not-too-distant future.

Since we're producing solely epub 3-books here nowadays, we've reprogrammed our in-house tools to cather for this, but - are there any available tools except for Jutoh, that do this kind of thing?

Also, I've prepared an epub 3, that passes most of the ACE-tests. (It's a sample of an edition of a Maupassant short-story collection (in danish), and sort-of copyrighted, but since I myself is the copyright holder, and it will be free to download from our site from next monday, I guess it's allowable.) Feel free to play around with it - comments are welcome, I don't think it's in any way perfect …

9788779795839-guy_de_maupassant-en_pariserborgers_sondagePROEVE.epub

Regards,

Kim

[JSWolf]

The problem is that most programs, apps, and Readers do not do ePub 3. iBooks does ePub 3. You should test these ePub with ADE 2.0.1to see how well or not they work and if they don't work well enough, then when they are sold, you'll need to specify that they require an ePub 3 compatible program.

There are so many ePub 3 (that should be ePub 2) that when an ePub that needs an ePub 3 reader comes along, people don't know and if they buy, they can end up with a book they cannot use or cannot use well enough.

[elibrarian]

Originally Posted by JSWolf

The problem is that most programs, apps, and Readers do not do ePub 3. iBooks does ePub 3. You should test these ePub with ADE 2.0.1to see how well or not they work and if they don't work well enough, then when they are sold, you'll need to specify that they require an ePub 3 compatible program.

There are so many ePub 3 (that should be ePub 2) that when an ePub that needs an ePub 3 reader comes along, people don't know and if they buy, they can end up with a book they cannot use or cannot use well enough.

I kinda expected this coming (You're quick!) But "stick to epub 2" is not an answer to my question, so I won't comment this further.

Regards

Kim

[Toxaris]

I would think that Sigil would not have any issues with those tags. You would have to do that by hand though, or perhaps someone can/will build a plugin for that.

[JSWolf]

Originally Posted by elibrarian

I kinda expected this coming (You're quick!) But "stick to epub 2" is not an answer to my question, so I won't comment this further.

Regards

Kim

I never said to stick to ePub 2. I said that there are too many ePub that are ePub 3 that should be ePub 2 because they don't use any specific ePub 3 features. This means that when a true ePub 3 is for sale, people don't know it is a real ePub 3 and it might not work with their setup.

So when you sell this ePub 3, you'll have to specify that it is a real ePub 3.

And yes, Sigil should do what you want as previously said.

[elibrarian]

Originally Posted by JSWolf

I never said to stick to ePub 2. I said that there are too many ePub that are ePub 3 that should be ePub 2 because they don't use any specific ePub 3 features. This means that when a true ePub 3 is for sale, people don't know it is a real ePub 3 and it might not work with their setup.

So when you sell this ePub 3, you'll have to specify that it is a real ePub 3.

It's not an issue. ADE 2 and 3 just ignores the accessibility tags and shows the epub 3 just fine.

Originally Posted by JSWolf

And yes, Sigil should do what you want as previously said.

At present Sigil doesn't do anything in the way of accessibility tagging. But on the other hand, it doesn't touch them in any way - destructive or not - either. (To clarify: We use Notetab scripting to produce the epub, and Sigil to polish and check it before publishing).

Regards,

Kim

[Doitsu]

Originally Posted by elibrarian

I guess that the ebook-distributors will demand new epubs to pass this as well as epubcheck in some not-too-distant future.

I looked into this and it seems the minimum requirements for reflowable epub3 books without scripts and multimedia files aren't actually that hard. All I had to do is:

1. Add the following section to the .opf metadata section:

<meta property="schema:accessibilitySummary">This publication conforms to WCAG 2.0 AA.</meta>
<meta property="schema:accessMode">textual</meta>
<meta property="schema:accessMode">visual</meta>
<meta property="schema:accessModeSufficient">textual</meta>
<meta property="schema:accessibilityFeature">structuralNavigation</meta

2. Add language attributes to all <html> tags.

3. Add chapter titles to empty <title></title> tags.

4. Add descriptions to empty <img> tag alt attributes.

5. Add role attributes to all tags with epub:type attributes that have an ARIA role equivalent.

Originally Posted by elibrarian

Since we're producing solely epub 3-books here nowadays, we've reprogrammed our in-house tools to cather for this

Does your in-house tool implement any other fixes that I've missed?

Originally Posted by elibrarian

Also, I've prepared an epub 3, that passes most of the ACE-tests.

I've updated a very simple MR epub3 book to make it WCAG 2.0 AA compatible. (Its only epub3-only feature is page numbers.)

[elibrarian]

Originally Posted by Doitsu

Does your in-house tool implement any other fixes that I've missed?

I don't think so. We haven't yet implemented the ARIA-roles (anyway, thats only a "minor" "best practice" error according to ACE, more or less the same level as epubchecks "warnings")

Originally Posted by Doitsu

I've updated a very simple MR epub3 book to make it WCAG 2.0 AA compatible. (Its only epub3-only feature is page numbers.)

Well, the ARIA-roles and epub:type in itself are also epub-3 …

Regards,

Kim

[KevinH]

So how could Sigil itself or a Sigil plugin help? Adding lang tags to every html tag in an automated manner seems easy to do. Adding a fixed set of metadata would be as well. Auto filling in title tags from the nav is also doable in an automated manner. A human would have to add the proper image alt tags.

Making it easier to mark tags with epub:type semantic info would be useful, I think. What else could Sigil do to make this easier?

[jhowell]

Originally Posted by KevinH

A human would have to add the proper image alt tags.

Or something like this: Google’s Image Captioning AI Can Describe Photos with 94% Accuracy

[KevinH]

BTW, these changes could easily be added to the ePub3-itizer plugin as well.

[KevinH]

If it is opensource and accessible in python, we could make it an edit plugin in Sigil that checks each img tag for alt and if needed tries to get a good description.

Kevin

Originally Posted by jhowell

Or something like this: Google’s Image Captioning AI Can Describe Photos with 94% Accuracy

[j.p.s]

Originally Posted by KevinH

If it is opensource and accessible in python, we could make it an edit plugin in Sigil that checks each img tag for alt and if needed tries to get a good description.

Kevin

It runs under tensorflow, which pretty much wants everything to be in python, but they also talk about how fast it is with some fancy nvidia GPU, so it probably takes a lot longer without a GPU. I doubt you want tensorflow to be a sigil dependency, so you probably want to define a plain text format for a file of image descriptions that sigil can insert as alt tags in the epub.

[Doitsu]

Originally Posted by KevinH

BTW, these changes could easily be added to the ePub3-itizer plugin as well.

IMHO, it'd be a good idea to add the following items to the ePub3-itizer plugin:

1. Add lang and xml:lang language attributes to all <html> tags.
2. Add NCX heading information to all blank <title> tags.
3. Add blank alt attributes to all <img> tags without alt attributes. (BTW, epubs without alt attributes will fail epubcheck.)
4. Add role attributes at least to the following default html template items:

<nav role="doc-toc" epub:type="toc" id="toc">
<nav role="doc-pagelist" epub:type="page-list" id="page-list" hidden="">
<nav role="directory" epub:type="landmarks" id="landmarks" hidden="">

5. Add the following boilerplate metadata section that should be applicable to most text-only reflowable epubs, as long as all <img> tags have alt attributes with image descriptions:

<meta property="schema:accessibilitySummary">This publication conforms to WCAG 2.0 AA.</meta>
<meta property="schema:accessMode">textual</meta>
<meta property="schema:accessMode">visual</meta>
<meta property="schema:accessModeSufficient">textual</meta>
<meta property="schema:accessibilityFeature">structuralNavigation</meta>

(Users will, of course, need to customize these values, if their books contain a large number of images or other items that affect the accessibility of the ebook, e.g. MathML sections.)

Originally Posted by KevinH

If it is opensource and accessible in python, we could make it an edit plugin in Sigil that checks each img tag for alt and if needed tries to get a good description.

AFAIK, the framework is opensource, but the training data isn't. And without the training data the framework probably isn't that useful.

[DiapDealer]

Could we make it a choice via the plugin's json file? Making epubs accessibility compliant is a noble endeavor, but i don't really want to add all of that for my own personal library (which I tend to convert to epub3 via the plugin).