Meeting minutes
<wendyreid> date: 2024-01-22
Wendy: Busy week last week so I was unable to get editorial work complete. New computer headaches ;)
… Today discuss "How do you make reflowable fixed layout books?" how to get content creators and Reading systems on board with this idea
How do you make reflowable fixed layout books?
SueNeu: Can we review the models we have looked at so far?
<wendyreid> w3c/
wendyreid: lot of discussion on various technices, this is a summary of them all.
… using Fallbacks, reflowable. multiple renditions, hybrid (fixed & reflow) charts fixed layout, for example.
… , guided navigation. Proposal visual to textual
… multiple style sheets fixed layout & reflowable. add a flag to OPF that it can convert toggle to use other style sheet.
… region based nav i.e. SVG
Using DPUB ARIA, could infer some of the properties of the text, ie. doc-frontmatter, doc-footnote how that should be rendered.
gpellegrino: Adrian, and others an approach of Reading Systems, enable RS to programatically construct reflowable out of fixed layout.
… PDF has this idea ISO standard enables content creator if they make the PDF in one way it can be rendered instead in HTML.
… a French company has a FL on one side and reflowable (font size colors etc. on the other)
<CircularKen_> Adobe PDF method is 'Liquid Mode' https://
gpellegrino: if we had a spec would be important to facility this conversion between FL and Reflowable.
<gpellegrino> https://
gpellegrino: Derivation PDF HTML spec.
CircularKen: I wrote an explainer visual rendering.
… I managed to create an example discarding the positioning span tags "ignoring"
… remove the FL declaration in the OPF, the Reading systems don't allow reflow.
… viewpoint media query was "a way"
… there is some CSS to discard the fixed layout positional tags on the spans. images are described would need to be there.
… could be done quite easily. I will find the link.
<CircularKen_> https://
SueNeu: "End users want to adapt to small devices.
… PDF document has this line which I like
gpellegrino: output transforming FL to reflowable, / reflowable view of FL. output can be achieved less on CSS / media query spans, but reading the actual document new on the fly in HTML.
<SueNeu> "End users..." line is last sentence of the first paragraph in this discussion about deriving html from PDFs https://
gpellegrino: works only in HTML, but with SVG could be a problem. Large images another issue.
… static content when you have JavaScript will be tricky with a reflowable view that is interactive.
… majority of FL is static so this could work.
CircularKen_: looking at a page without looking at the code, we are trying to ascertain the reading order based on position of text on the page.
gpellegrino: we should leverage what we have. 100% we need content creators to create good FL books reading order / images with alt etc.
… can benefit from DPUB ARIA roles, descriptions of images.
<wendyreid> w3c/
wendyreid: issue 60 in repo we should include this. using DPUB ARIA
… existing HTML EPUB practices, explain what would we want RS to do with the DPUB ARIA semantic info. especially if we bring to the broader CG.
Hadrien: same thing as TTS, you need to walk through the DOM. you break it down into utterances, use ways to cutting things down to sentence or less to send to tts utterance engine. extract useful info like alt. based on that you recreate a view. it raises questions like skipability. maybe a best practice doc. second best practice doc how to create the reflowable view. how to extract text.
… major requirement is needed here.
wendyreid: tts at last summit, most don't read image descriptions with read aloud. there is no TTS best practices not sure where this fits as far as scope.
… lots of websites have this TTS (Read Aloud) listen to an article for example.
… a publishing problem, publishing is asked to implement this widely / legally. what should be the best practice for this tts/read aloud.
… maybe something to raise with the wider CG
Hadrien: also similar with Browsers.
… TTS in Chrome, android also is moving on this.
… if we document something, we would love to know how you do it / compare with others. EPUB reading system would be a good start. may be difficult FF/ Microsoft / Apple /Google agree or listen to us.
wendyreid: 2 strings of work. Best practices TTS / reflow presentation (how to walk the dom) Then the techniques how to do this in an EPUB. How to construct your FL content in the best way to facilitate that Best practices work.
wolfgang: next CG mid Feb.
… couldn't we say 2 issues, what can you do in content structure to facilitate to FXL to reflowable, then the walking the dom how to speak this etc.
wendyreid: New issue TTS / walk the DOM to produce the derivation of content to produce a "ultimate text renderer"
… who wants to take on the "ultimate text processor" ? new name??
… then we can present both of them at the next CG meeting in a few weeks.
Laurent: when Gautier proposed this, display transform Options.
… is this what we are talking about how to transform FXL to raw text?
wendyreid: Yes I didn't want to put it in the main document as it is not ready mainstream / common. Don't want to give the impression that this is a common feature / implemented yet etc.
… until we settle on how this will work in details. can be added in the future.
SueNeu: ultimate renderer / tts based on DOM, current TTS does an accurate job for my text, 98% accurate tts might need to be better, publishers need to do better before we rely on this.
Laurent: what do you mean not 100% accurate.
SueNeu: we are stuck with the voice with the reading system, editors want a better voice, need to make sure reading order for sure, the TTS will take ALL CAPS read letter by letter, etc.
Laurent: reading order is up to the publisher, 2nd support the developer of the RS these fixed layout must be transformed satisfactory.
… 3rd is the voice, we use the operating system for now, they will evolve. publishers has some responsibility but the rest will evolve.
Hadrien: when things are read letter by letter the content is badly authored, something in DOM. the TTS can be smarter, should not be separated etc.
… voices very complex. tested 500 different voices / 5 languages / browsers / OS etc. I documented 100 of them nice voices for our web reader. nothing documented
… depending on the OS there are decent voices, and others have very low quality languages, may be dependent on what you install others you may not have an option. There are things RS can do to make this better.
… RS should you read page breaks, semantic info of ARIA roles, some may need it some may not. need option settings.
… user perspective may not be ideal, they don't know if its a content, configuration problem, etc.
… why I think we need to break this down to smaller problems.
CharlesL: The thing that was mentioned about all caps, I've seen where it's all fine, or when the screen reader or TTS reads it by letter
… I'm recommending in GCA to use CSS to make the content appear in all caps
… when the word could have different pronounciations depending on context, that is something that is improving
… there's cases where they just pronounce incorrectly
wendyreid: CSS Speech?
wendyreid: EPUB attempted to tackle this PLS how to help with pronunciation.
… other groups SSML etc.
… CSS Speech is the closest and this should help moving forward as this becomes more mainstream.
… 2 topics, can I have a volunteer for the Rendition Issue.
Hadrien: I can do this.
wendyreid: Thanks Hadrien, no meeting next week we have the Accessibility Summit / AB meetings.
SueNeu: a friend makes a lot of media overlays promises comments on our document and I will add them to the github discussion.