Introductions...
... Mat King
... Matthew
... Ian
... Janina
... Martin Robertson
... Gottfried
... Steve Lee
... Joanie Diggs
Janina: We will have people from the Math community joining us at 11am.
... We want to spend this hour to prepare for this meeting.
Joanie: Where are we and what are the next steps?
... Peter, the co-chair of the MathML community group, disagrees to continue with MathML.
Mat: Which other technologies are being exploited today?
Joanie: We do not have consensus on that.
Discussion on whether MathPlayer is still alive and being developed.
clapierre +1
Gottfried Joanie: VoiceOver was supposed to support MathML:
... I want the ORCA screen reader to support it, but i am not a Math person.
... Charles Lapiere introducing himself - joined late
Charles: I can give you some background on the Math discussion.
... Peter was pushing MathML, but got pushback.
... MathML is based on XML.
... Peter tried to get support for MathML into the browsers.
... Peter has now decided to turn around and explit other technologies to get math working on the web.
... Workshop "MathSummit" in the past summer. Thoughts about Math being brailled and spoken.
... Using ARIA, made some hacks.
Joanie: We have some things that would use ARIA without making hacks. We just need to change ARIA a bit.
Janina: Math specifics: equations, graphs, angles, shapes. Part of it is SVG, and part of it is a natural fit for ARIA.
.. Symbols are in Unicode, so we need to nest and group.
Charles: There is presentational MathML and content MathML. Most of the work happens in presentational MathML - millions of works from publishers.
... Content MathML is about structuring and meaning of the symbols.
... Publishers are now stripping MathML off their books, and replacing it by images, even without text descriptions.
... Rumors that the MathML wg might get reborn.
... Neil could be leading this wg.
... I had a discussion with him on Friday.
... Rendering of MathML is good in general. But publishers have some performance issues.
Mat: Publishers are required to produce accessible books.
Charles: Some publishers are putting alt text on images, using Benetech MathML cloud. MathJax will take the MathML and create alt-text for it.
... We have an accessible math test book with various techniques. All will go into epubtest.org
... Goal: SVG image, MathML hidden off screen. MathJax could be employed by JavaScript. AT can get MathML if wanted.
Ian: From my limited experience, MathML is a bit like SVG. With simple things, works well. But gets complicated with more complex things. Not much tooling around.
Irfan: When hiding alt-text, what do braille users do?
Charles: It is all in the hidden MathML that is visible to AT for using MathJax or browser support.
... Caveat is that you lose low-vision support for highlighting.
... But currently not much supported by AT anyway.
... Better support would be: Math rendered onscreen, and have semantics embedded in it.
Janina: That's like reinventing MathML for structure.
... That's why Neil wants to re-start MathML wg
Charles: There is a minority saying that MathML is dead.
Janina: There is not necessarily just one approach.
joanie q+
Gottfried q+ ask on MathJax
... Some people are looking at corner cases (just 1%) where MathML would not work.
... Neil wants to focus on the 99% where it works.
... In the end, we may also target the corner cases.
... Neil really wants to get the MathML wg going again.
... Neil says we will still need SVG support and other technologies that Peter wants to use.
Joanie: What do we agree on? (1) We want a consistent user experience. (2) Speak it correctly, and braille it correctly. A screen reader does not necessarily have to understand the math semantics.
Charles: We cannot add all the different semantics.
Joanie: Does the tf have consensus on what approach to take?
Janina: We agree on a good user experience.
... But problem is not just screen reader users. We have to think wider, including digital publishing. I guess it would push it toward authoring.
Gottfried q+ to mention AMAC's abandoning of MathML
Charles: There are data visualization groups working on this too.
Janina: We need to slice the problem space, and get one piece accessible, then next, and so on.
joanie We already (hopefully) agree: The goal is to get a consistent user experience independent of what host language is used to render content.
joanie Question 1: Do ATs need to know the underlying semantics to provide support, or do we want to put $THINGS into place so that ATs magically do the right thing automatically?
Janina: if we put it on AT, then every AT will do it differently -> no consistency.
Mat: For ARIA we said, there would be no normative requirements on AT.
... There could be another layer feeding AT.
Joanie: Ideally, we have MathML rendering everything. Peter has this SVG thing with choices. To get this totally accessible, you need a CSS-hidden HTML table, etc.
Mat: For a given user, and a given technology, regardless of coding...
Gottfried: A "consistent user experience" does not mean it is the same for every user. Personalization is important.
joanie New Version: We already (hopefully) agree: For a given user of a given AT, there should be a consistent user experience independent of what host language is used to render content.
Gottfried: What is MathJax?
Charles: MathJax takes MathML, renders it, and allows AT to navigate in it.
... There is a speech rule engine with a dictionary.
Ian: It is like a polyfill for MathML.
Gottfried: AMAC has abandoned MathML production. Temporary alternatives: Enhanced Tagged PDF (ETPDF) files, Long Description Microsoft Word Doc (DOCLD) files, and EPUB files.
... Does the Math community group have producers of alternative learning materials with them?
Charles: Many people work with Latex.
Gottfried: I could contact AMAC and ask for their support on this.
Ian: Is high school text rendering the user case for us, or academic rendering?
Ian: Maybe it is fine to start things with the consistent rendering?
Joanie: Goal is to find out what the next steps are. We need to find out if these things need to be squeezed into ARIA 1.2.
IanPouncey To clarify, is the best approach to achieve adoption to have consistent rendering of basic math but lacking ability to describe high level math, or is it more important to focus on describing high level math possibly at the cost of consistent rendering (at least initially).
Janina: Knowledge-domain accessibility.
Joanie: Role description could help to let SRs "magically" speak the right things.
... Role description is a case where we need a braille strang in addition to the spoken strang.
Charlie: Benetech is doing a Math editor. Need braille input.
joanie https://github.com/w3c/aria/issues/771
joanie Consider addition of a property for input controls to provide untranslated Unicode braille key input events
Joanie: Should ARIA wg in their own meetings try to tackle this?
... With the Math community group, we need to figure out what the next steps are?
Janina: Use metadata to declare things, and preload stuff.
Mat: We need to define the most important problems.
... Problem of consistent user experience.
Joanie: The user should not have to know what the underlying markup is.
... SVG is ideal for rendering according to the author's intentions.
Joanie: SVG elements have to be exposed to AT.
joanie New Version: We already (hopefully) agree: For a given user of a given AT, there should be a consistent user experience independent of what host language is used to render content.
joanie Question 1: Do ATs need to know the underlying semantics to provide support, or do we want to put $THINGS into place so that ATs magically do the right thing automatically?
Janina: Almost okay with that. Needs to be compelling.
Gottfried: Should add "personalized user experience".
Ian: "Consistent, given your choice of user experience"
joanie New Version: We already (hopefully) agree: For a given user of a given AT, there should be a consistent, compelling, and robust user experience ***independent of what host language is used to render content.***
Mat: NVDA users vary in their settings. That's personalization.
... This is different from the personalization semantics in the personalization wg.
Janina: "Compelling" means that as a reader you can discuss this with your classmates.
joanie Question 1: Do ATs need to know the underlying semantics to provide support, or do we want to put $THINGS into place so that ATs magically do the right thing automatically?
Joanie: We need to split work up so each individual group can go off an do their work.
Mat: As an AT developer, you are not going to develop support for half a dozen math description formats.
Joanie: Like "btn" (AT knows it is "button").
Janina: We are not done when we get math done. We still have chemistry, linguistics, inter-linear content. We need to do knowledge domain.
Joanie: AT do not need to know knowledge domain semantics.
MichaelC just noticed rrsagent is not in channel, minutes not logged
Janina: Somebody with knowledge domain expertise needs to create these mappings.
Mat: Accessibility tree strictly becomes a rendering engine.
Ian: Will this be available in browsers, so that AT can support it?
Joanie: Implementing support for ARIA roles is easy.
... And it is easy to expose these roles in the a11y tree.
Ian: Then we are back to the MathJax approach, i.e. somebody is doing this for the browser.
Mat: How can the semantics be expressed in an a11y tree?
Joanie: Already by MathML.
Gottfried: We do not want AT to be required to understand the math semantics, and we do not want to prevent AT from understanding the semantics (important for cog a11y).
Mat: Is ARIA going to be explode into something really big?
Janina: Yes, you can map stuff easily.
Joanie: ARIA is a hammer and a screw-driver and a wretch.
<arnog> hello
<clapierre> tackbot, start meeting
<clapierre> trackbot, start meeting
<mck> js: Lots of knowledge domains have their own ways of marking up content
<mck> e.g., math, chemistry, music, etc. need generalized principles that can be extracted to other areas.
<joanie> For a given user of a given AT, there should be a consistent, compelling, and robust user experience independent of what host language is used to render content.
<IanPouncey> https://github.com/w3c/apa/issues/9
<mck> jd: we hope we have consensus on this statement.
<mck> jd: It should not matter what the underlieing markup is.
<mck> js: Many diff kinds of AT and disability, we want support for all so they can have intelligent and meaningful conversation with one another.
<pkra> +1
<joanie> Do ATs need to know the underlying semantics to provide support, or do we want to put $THINGS into place so that ATs magically do the right thing automatically?
<mck> jd: Orca doesn't have math player, for instance. Does a screen reader need to provide custom nav between numerator and denominator ... if yes, then screen reader needs to know which element is numerator.
<mck> If sr just needs to be able speak or braille a numerator and does not need to know that element is numerator, then sr does not need to know semantincs.
<mck> Peter: Idea of community group is to go bottom up rathe than top down.
<mck> Start with concrete problems rather than abstract issues.
<mck> Had workshop in san jose where brought togther tool devs, publishers, sr devs, etc.... last summer.
<mck> Volker: Answer about semantics ...
<mck> It isnot the job of the AT to know the semantics. How could they?
<mck> Creating dictionary of math roles has been tried. It has not worked.
<mck> Peter: Equation presentation is a small part of math domain.
<volker> Summary: Science is open ended and grows everyday.
<volker> AT does not need to know all the semantics. But needs an open way authors can add their content.
<mck> Solutions for math equations may be different from equation layout in other disciplines, e.g. physics.
<mck> gz: Author does not requirements of readers.
<mck> correction: does not know requirements of readers.
<mck> AT may need to build a bridge for specific requirements of readers, e.g., to help with comprehension.
<mck> Arno: for simple thigs it might be possible for AT to provide interpretation.
<mck> This could be possible for fractions, but fraction presentation can mean many different things. AT cannot have all that knowledge.
<mck> AT may read something as a matrix ut it is actually binomial coefficient.
<volker> Different fraction notations https://github.com/w3c/mathonwebpages/wiki/%5Ba11y-TF%5D-ambiguous-notation
<mck> Peter: AT track record for interpreting sematnics is not great. There is a trust issue here.
<mck> Solve underlieing problems of authors for creating good experience, solve broader range of accessibility issues.
<mck> Equation layout is very print centric
<mck> We have hacked print to do things like using different fonts for diff semantics.
<mck> GK: Publishers want a digital product that is visually appealing.
<mck> If it does not look good, it doesn't get sold by publishers.
<mck> js: No AT has the expertise to understand all these knowledge domains.
<mck> If we have tooling for good mapping, we could address these issues.
<mck> Names spaces can be problematic.
<mck> Dani: could fine to write a solution for some specific semantics like fractions and square roots.
<mck> But, we need more general solution for the complete domain.
<slightlyoff> (I realize I'm an observer and not entitled to participate, so please dispose as chair sees fit)
<mck> AT should not have to know the semantics ... we express the semantics with something like labels.
<Zakim> joanie, you wanted to suggest at some point today it would be nice to enumerate those needs (in addition to the three issues we have open against aria)
<mck> But, we could let AT know about some limitted small set of semantics.
<mck> Alex: The idea of general extensability is consistent with with is recommended in other areas.
<mck> It has been suggested that ARIA be more extensible. Extends what HTML can do.
<arnog> +1
<joanie> https://github.com/w3c/aria/issues/425: Remove children-presentational true from math role
<mck> jd: Issues filed for ARIA
<pkra> https://github.com/w3c/mathonwebpages/wiki/%5Ba11y-TF%5D-TPAC-2018-preparations
<mck> Presentional children false is a no brainer for ARIA group.
<mck> Peter: concerned about existing content that uses role math that expects children to be hidden.
<mck> jd: Safari may e doing that bug gecko does not ... she thinks.
<mck> need to check.
<mck> Volker: Navigating ... we have tried using aria-owns but it is awkward.
<pkra> I pasted the link above.
<mck> Rather than fixed roles, need some ways of providing summaries or descriptions.
<mck> Implementation consistency coul be one of the problems.
<mck> Arno: Implementation could e part of the problem.
<mck> In some cases, the spec itself could be interfering.
<mck> In some implementations, label and description are combined.
<mck> While you may only want the description.
<mck> It goes back to giving the author control over presentation.
<mck> AT presentation.
<mck> Peter: We have some time with CSS group later today.
<mck> Provided link with issues that have wish lists.
<clapierre> https://github.com/w3c/mathonwebpages/wiki/%5Ba11y-TF%5D-TPAC-2018-preparations
<mck> We see some very basic problems we would like to tackle. We want these to be starting points for long term solutions.
<mck> Math is a niche, and there are limited resources.
<arnog> It is difficult to know where the issues lie: in AT, in spec, in content produced...
<Zakim> joanie, you wanted to talk about braille and aria-owns
<arnog> being able to work more closely with experts in the WG would help narrow down where the issues are
<mck> jd: ARIA wg needs to focus on specialized formats, right?
<mck> We see this is something we need to have in the wg.
<mhakkinen> +1 to ARIA addressing braille as a separate "label"
<mck> Question is matter of priority in ARIA.
<mhakkinen> +q
<mck> gk: Are we thinking this applied to svg elements?
<mck> jd: I think it would independent. You can use aria on mathML, svg, html, etc.
<mck> Daniel: When we think Peter: if I want to give description for voiceing, then that has a very different shape from nimmeth braille
<mck> Daniel: if we have 10x10 matirx, what kind of UX do we want?
<mck> Not just speak out all 100 numbers. Or drill inside.
<mck> Peter: May need to customize exploration techniques.
<mck> MH: Big challenges with educational assessments, is to be able to present in both spoken and braille.
<mck> Speech we have asked to address in pronunciation tf
<mck> Arno: May need multiple braille expressions of same element.
<mck> Nimmeth, UEB, etc.
<mck> In order to correct Nemeth translation, need to have semantic understanding.
<mck> At authoring level, it is easier to provide correct spoken presentation, authors can do a better job of translating to nimmeth.
<mck> Almost impossible for AT to have semantic understanding; authors have that.
<mck> Volker: We are pushing speech out now with aria live regions.
<Zakim> joanie, you wanted to ask specific question about aria-owns versus new-nav-thang
<mck> jd: Would aria-owns work if all UAs reordered tree in same way.
<mck> Would a combination of aria-owns and aria-flowsto help.
<mck> aria wg needs to know what is most needed.
<mck> Not sure problems are lousy authoring or inconsistent implementation.
<arnog> \me asks joanie: which github issue?
<mck> js: There are many different types of needs in presentation. We cannot be aware of all of them.
<arnog> :)
<mck> Eagk: We know AT cannot interpret all these semantics. Cannot be done.
<mck> Providing a small set of features, such as exploration and navigation methods, braille output features, next stpe to define the AT presentation requirements.
<arnog> +1
<mck> Start with some AT requirements and then get at least one implementation.
<mck> gz: is there a list of use cases we can start from.
<mck> Peter: we can get the use cases.
<mck> jd: Shortly after san jose meeting, I suggested joint task force. There was push back.
<mck> Another option is to expand aria membership.
<mck> We could have some invted experts.
<mck> Peter: I think there is enough interest to make working with ARIA wg work.
<volker> Some use cases for equation systems/tables http://zorkow.github.io/speech-rule-engine/resources/www/
<volker> @clapierre Yes.
<mck> jd:summary: we will invite math people to aria and add to aria agendas as appropriate.
<clapierre> https://github.com/w3c/mathonwebpages/wiki/%5Ba11y-TF%5D-ambiguous-notation
<LisaSeemanKestenbaum> i am on the phone
<LisaSeemanKestenbaum> anyone there?
<lisaS> anyone there?
<lisaS> im on the webex
<lisaS> i am in the webex "apa at tpac" is that right?
<IanPouncey> Lisa: we'll start webex soon.
<lisaS> thanks
<lisaS> webex reminder
<lisaS> thank u!
<lisaS> (if it is too hard just let me know)
<lisaS> (I can always take a nap :)
<STeve> scribe:STeve
<lisaS> thanks
<JF> Comparison Page: https://github.com/w3c/personalization-semantics/wiki/Comparison-of-ways-to-use-vocabulary-in-content
LS explain our work and describe mechanisms to attach the metadata we want to choose from
<JF> Taxonomy (ies) Personalization Semantics Content Module 1.0(https://w3c.github.io/personalization-semantics/content/index.html) Personalization Help and Support 1.0(https://w3c.github.io/personalization-semantics/help/index.html) Personalization Tools 1.0(https://w3c.github.io/personalization-semantics/tools/index.html) Personalization Semantics Explainer 1.0(https://www.w3.org/TR/personalization-semantics-1.0/)
rrsagent: make minutes
<lisaS> https://www.w3.org/wiki/TPAC/2018/SessionIdeas#Personalization_-_How_Should_We_Implement.3F
CL will do the introductions to the WPG
JF suggests we indicate all formats we looked at but not details as we only have 15 min
The 3 items are thosew with with double asterisks in the comparison URL above
LS new attribute with addition mention in survey with is missing from the options (should be 4)
<lisaS> <h2>3 Attributes</h2>
<lisaS> <label for="given" aui="simplification:critical; field:givenName" easylang="First name">Given name:</label>
SL wonders if covered by one of the other 3?
LS says it's not
Janina: we have very limited time to address this
IP suggested suggested giving WP three to chosse from will be opening a rat hole and better to listen to them
MC says they will look for least damaging to HTML and easy to sell to WhatWG
LS points out the comparison document does not provide details of all the selected formats
We are moving to the webplat meeting minutes will be imported from their
join: webplat
<clapierre> ScribeNick: clapierre
<Irfan> scribe: Charles
Janina: synchronizing, possibly more, maybe a standard proceedure.
we look for, sense of what the FPWD is looking for.
as something comes to finalization we come to good place, anything coming from you folks that you won't miss the basics. we will be looking for edge cases, I think.
there is only 1 spec?
Avneesh: Yes
<romain> Web Publications spec:
<romain> https://w3c.github.io/wpub/
you will inform us if you find anything?
Janina: Yes
Micheal, we review on a monthly status, but as it gets close to finalization send us a request for wide review.
If you have any questions about the a11y of your spec please ask questions. Via public email list
Avneesh: sometime mid next year it will go to CR
MichaelC: so then a few months before go to Wide review
We can skip personalization as you are already working with Charles.
George: epub 3.2 has been approved by the Community Group but now needs support from the Business group.
Avneesh: Sync media, mathml, iso summisions.
<romain> scribenick: romain
janina: a lot of it comes from
COGA
... what kind of stuff do you want to markup content with
clapierre: the Personalization TF is now part of AP (was part of ARIA)
<Roy> rrsagent: make minutes
clapierre: came out of COGA and
IndieUI groups
... the goal is to enable personalization of Web content to
personal user needs
... add and adapt content to more scennarios
... sometimes users need personalization of text or content,
symbols, number-free, additional help, etc
... UA or other tech can use those semantics to augment or
adapt the concept based on user preferences
... e.g. "9/10 people prefer coffee over tea" can be adapated
to "almost all…"
<JF> Taxonomy (ies) Personalization Semantics Content Module 1.0 (https://w3c.github.io/personalization-semantics/content/index.html) Personalization Help and Support 1.0 (https://w3c.github.io/personalization-semantics/help/index.html) Personalization Tools 1.0 (https://w3c.github.io/personalization-semantics/tools/index.html) Personalization Semantics Explainer 1.0 (https://www.w3.org/TR/personalization-semantics-1.0/)
clapierre: for digital
publishing, we want to support personalization as well
... be able to personalize symbols is a good use case for
instance
George: the various HTML element,
like headings, can have a CSS representation with symbols so
that headings appear in a certain way
... also the DPUB ARIA vocab, which could also associate with
some symbols
... like the warning element (skull and crossbones), etc
... seems pretty doable without major modification to the
content
clapierre: some of the
personalization is also related to typography, color
combination
... how content should be presented to you
... it's a traditional form of personalization
George: is that content or UA?
JF: probably a combination of
both
... the personalization can set the condition in the code, that
the UA can use
... e.g. the UA would map a token value to a particular symbol
(skull and crossbones)
... the visual expression is left to the UA
mark: it's a bit similar to
responsive design
... you'll have people questioning the change of look and
feel
IanPouncey: on a case by case
basis you can do that with stylesheets
... it's not that it can't be done
Avneesh: do we need to do anything more from the PWG side?
clapierre: you need to tell me your use cases so that we can make sure to incorporate it
janina: and share the vocab, so that we can map it
JF: see the links I pasted earlier in the IRC
clapierre: we don't know yet the
mechanics, what the syntax will look like
... it's still up to discussion and debate
... it'll probably look like 2 or 3 attributes
JF: 3 types: values from a token list, URI, and strings of text
clapierre: simplified language
might have a paragraph, multiple sentences
... it might be better to have that as a single attribute
... more concise things can use a token list
George: is that a substitution?
clapierre: correct
George: where would that stuff live?
JF: in the document, or at a URI
if it's really long
... for instance to simplify a long legal text
... the actual final solution would be left to the author
clapierre: that could even go into image descriptions
JF: to the average user it may
not be obvious, exposing it may depend on browser
extensions
... the goal is to set the condition in HTML so that machines
can manipulate the data
IanPouncey: similarly to ARIA attributes: it's there and anyone can process it
janina: sometimes the
requirements compete
... you can't always have all requirements for all users at the
same time
IanPouncey: like "easy lang" can
mean different things for different people
... it might not be the right name
janina: it would also be useful
if a 3d party could do this markup, like a university doing
specialized content
... although it raises licensing and copyright issues
... I expect you folks might have a twist on this we haven't
thought about
George: keep on eye on what's
done in applications like Edge
... what they're doing is amazing: more spacing between the
lines, font combinations, color personalization, etc
... all this can happen on the fly with the content
... simplified language stuff is very interesting and can go
into AI in the future
clapierre: it's the "immersive reading" project that MS has done and is now part of Edge
George: it usually means that you have audio and text, or audio and Braille, at the same time
<clapierre> scribeNick: clapierre
<MichaelC> Knowledge domain accessibility
Janina: ARIA asked to help MathOnWeb CG
we had a meeting with them this morning, lot have direct experience making math accessible. impressive Group
We won't be finished when we get math working better, and probably not get all of it done as there are new math etc.
this will continue to change so we will never finish, another way its not just mathematics, chemistry, biology, etc…
STEM is an obvious candidate, linguistics is another non english text which you need pronounced / handled correctly
ancient history could get referenced
Standard study tool for ancient languages / sacrid text if you just read the english and looking at the original hebrew arabic, would get some sense with 2 streams of text.
Music, on a timeline and good at sync. multiple levels 60-70 parts
all kinds of ways this can play out, no one in AT will be able to do this, you need expertise in this knowledge domain.
Can we looking at several disciplines that we can come up with a framework, grouped effectively and used correctly and links correctly.
no one approach will be the only solution, there there are multiple ways to making content accessible.
Irfan, will be starting a new TF on pronunciation.
different knowledge domains, we may need a certain widget we may not have, some way to layout a vector to do this logically.
we may discover we need to invent something.
George: Charles and I are doing a presentation at AHG, one thing we have is a principle with these domain areas, that folks with learning disabilities don't use AT, they cut/paste and put into their own tools.
people may want to enlarge it, hear it or feel braille, and navigate it.
some kind of tool/widget maybe it could fall under AT domain could find a thing to present it that way.
specialized tool SVG content. from math it may be a tree, matrix, if its chemistry could be circular, so we may need different ways to navigate this.
Seems like developing a system that could handle multiple domains, creating content mathML for that, mathjax producing the aria content for this kind of stuff.
I think this is great, I am worried about complexity and scope
Ian: talk about music/linguistics, string session vs. procussion, personalization if you consider string session being a way to personalize any potential solution
lot of similarity to achieve there.
Janina: you are taught to take a logical piece for music, you may want to also slow it down, may need to repitch it
we have all the pieces to do this, it is useful you can pipe that to your braille display.
This isn't restricted that you can personalize these scores.
Ian: thats why I am conflating this with personalization.
George: one of the obvious customization to change the pitch speed of how they are spoken.
What are we going to do with this?
What about domain knowledge work?
Janina: APA for now.
Joanie: ARIA is trying to just stick with ARIA.
until there is a new state/property or role.
George: best practices for what we think we know, Publisher / apa joint thing. we got the Math book with test examples and our goal is to get a recommendation for publishers for their math. It does pretty good job for blind people, but not ideal for LD.
What is this? TF, CG? something else?
Janina: we don't know yet, its just an issue #9 currently
APA is also licensed to do normative specs now?
S/?/.
George: Longman domain has really cool stuff
<MichaelC> Langfocus channel
great body of work.
Michael, highlighting parts of speech.
markup you might do and express.
not sure if it will be in 1809 windows or the next version with highlighting parts of speech.
<DanielWeck> https://www.onenote.com/learningtools
<DanielWeck> Example Microsoft Immersive Reading
<DanielWeck> (click try it now)
So what are we going to do? are we coming to a recommendation at TPAC?
Janina: we have 1 more discussion on this tomorrow @ 1pm
hoping to make this wider than just DPUB.
George: publishers want correct visual presentation but don't think they will have domain knowledge.
JohnF: If we build it they will come.
George: Yeah.
Marisa: text is in HTML5
we want this to work in the context of html5, we want to sync audio clips with text, DAISY did this and EPUB di with Media Overlays with SMIL
step 1: look at existing technologies, using SMIL, can ref. text in another doc, but no browser support, dev. have not shown love not actively maintained. it does exist though as a rec.
TTML2 is another option it can accomodate non text audio clip but can't ref external text easily.
<romain> use cases: https://github.com/w3c/sync-media-pub/blob/master/use-cases.md
<romain> additional requirements: https://github.com/w3c/sync-media-pub/blob/master/addl-reqs.md
<romain> potential tech solutions: https://github.com/w3c/sync-media-pub/blob/master/technology-selection.md
you could put text ref as an attribute, id, xlink but no semantic values. feels like a hack.
looked at webvtt similar problems could not have different granularity, must support nesting.
web annimations not declaritive language.
annotations is another option
so this leaves us with making our own.
Daniel: some of the drawbacks with SMILE was proposed by DAISY, EPUB3 with MO, over the years we have simplified it.
Media Overlays with audio and video, web people do not like xml
EPUB 2 was xhtml, EPUB3 is html5 and other technologies in stack, JSON etc.
started with Readium2 project. look to SMILE xml to lightweight, internal format based on JSON.
since then we have looked at bowwowing readium2, and will present to the Publishing WG our work tomorrow
tree Data structure, we want to map this tree structure with audio clips. 1hr mp3 we want to linc to text/images/ etc. whatever can be narrated.
Daniel: take full document that are fully structured, styled semantics and we want to be able to add in a non destructive way to sync an audio track too.
we are looking at taking web as it is now and sync audio to that.
take a video ttml captioning track within that file there will be how to render the captions
George: millions of these EPUBS all working fine with TTS which is fine, but we are talking about Narrated audio. We need a mechonism to put these two different tracks down text track and audio track and we need mechognism to tie these together.
John: you want the text to be the master?
Goeff: have you tried to strip off the xml?
Daniel: you still need the processing model
<irfan> scribe: Irfan
Daniel: Json is most practical
but this is encoded as pair value which is referencing
different node.
... we do it by fragmenting ID's usually. we are not changing
that but proposing a technology called CFI which is very useful
and publishing world wants it. It allows the selection of
characters.
DanielWeck: we are proposing to align it to open web platform called media fragment URI's. we are using the notion of hash (#).
Daniel: we are proposing the extension of CFI.
JF: have you approached to TTML group??
marisa: yes.
Daniel: we dont want to re-invent another syntax. two- we should be able to re-use the software. choosing TTML would mean, we wouldn't be able to re-use the TTML because there is no gain to use that.
<Zakim> MichaelC, you wanted to ask about coordination with other groups in W3C
<marisa> https://github.com/w3c/sync-media-pub/blob/master/technology-selection.md
MC: coordinating with TTML, do you have a sense that there is a risk to stepping on each other toes?
Marisa: that wont do it.
MC: I am being devil's advocate but need to make sure that we are not re-inventing anything.
gottfried: you are creating a more powerful tool than TTML. potentially it could replace TTML even when you have caption that is spoken right now which could be a problem. that kind of overkill of the usual patches.
Mark: I have been longing for 15 years to have some solutions.
gottfried: what was your use case?
mark: back in 2002, we were
thinking about what could we do with a multi media daisy
standards. how would different directors synchronize with same
scene.
... I think we have use cases at ETS no one got the authoring
platform. everyone has custom solution.
gottfried: this is a great usecase and I want you to continue to explore the idea to resolve the issue.
marisa: currently we dont have good representation to address this issue.
Avneesh: there are some use cases in publishing industry.
George: there is a use-case if the sign language teaching is 5-10 minutes, it works otherwise there are some complications.
gottfried: other use case is short video clips that goes for non-native speakers.
<clapierre> http://livinghistory.sanford.duke.edu/interviews/dana-priest/
george: we got the audio publishers and digital publishers and they will implement this solution.
<clapierre> link to work from Kim Patch and Rutherford Living History synced transcript to video
<DanielWeck> Past experiments: http://wam.inrialpes.fr/timesheets/annotations/
<DanielWeck> :)
marisa: tomorrow 11 we are going
though it in details.
... audio books is another group where we will be talking about
it.
daniel: we are constraint with
the limitations. we use the term overlay. in reality there has
to be some degree of editing in HTML markup .
... there is some degree of disruption because of various
limitations.
janina: we are short of time. lets move to another topic.
Avneesh: ISO EPUB
accessibility
... EPUB 3.01 is already published. that will become ISO
international standard soon.
... its a small document and we can adopt it according to ISO
standards. last date is dec 25th.
... possibly EPUB accessibility standards will become
international standards
<MichaelC> scribe: MichaelC
https://www.w3.org/WAI/APA/task-forces/pronunciation/
js: new TF, Irfan is facilitator
may interest Publishing
mh: ETS writes high-stakes tests for students to demonstrate knowledge or skill
US states pay for assessments and results
studends with learning disabilities or language learners use text-to-speech
many states have specific rules about pronunciation
of words, numbers, where to insert pause
currently, assessment vendors use various hacks to achieve this
we can have to deliver tests on a variety of platforms
getting correct pronunciation across that range is difficult
the hacks are pretty ugly
and often work on one system, but work embarassingly wrongly on others
we looked into this and wanted to use SSML
can author it, but difficult to get the content into HTML where readaloud tools can access it
have talked to W3C groups, AT vendors, browser vendors
haven´t licked it
we are considering JSON as a way to bring SSML-like behavior into HTML
browsers don´t have to know it´s there
jd: have to pass it along
mh: to AT; how that gets done is the question we still have open
jd: AT used to scrape the HTML; now many look at AAPI, so it has to turn up there
it´s easy to pass on
mh: readaloud tools don´t use AAPI
but we do agree, need to work with screen reader vendors
gk: readaloud is a consumer?
mh: yes
gz: so how will you represent pronunciation? International Phonetic Alphabet
mh: subset of SSML
IPA can work in that context
some implementations don´t support SSML
putting the speech properties into JSON means it doesn´t matter how it comes out
gz: who does conversion?
mh: AT
<scribe thinks saying AT must convert to a native feature for pronunciation rendering?>
mh: it´s light-weight
the TF will explore other models as well
CSS speech is an option, but hasn´t worked well so far
ia: no longer in development at Apple
gz: love the idea of a pronunciation style sheet that would modify a single piece of content
but guess not
gk: so how does it work? Markup in a span?
mh: the JSON can be in a data- attribute for an element
would like to have a way to centralize the JSON to avoid fragmentation and duplication
ia: many ways to attach the JSON to content
some of which allow multiple properties on one chunk
@@1: @@2
EPub3 has three mechanisms for pronunciation
mh: we looked at that, and discussed with IMS Global Learning Consortium
some of these technologies don´t get adopted
did get some into IMS standard
but still run into trouble getting it into the HTML
we now think we have a proposed solution mature enough to discuss in a formal task force
gk: what about pauses?
mh: there is a <break> in SSML, can add a time delay
can make as a break before or after element, or an inserted breaking element
we get questions about authoring difficulty
we think any new feature has authoring difficulty
would consider authoring guidance in developing
DanielWeck: : do think SSML better fit than CSS speech
the latter seems to be being downgraded
mh: CSS speech doesn´t have everything SSML has, so we want to focus on what solves it most comprehensively rather than support multiple formats
gk: I thought reading system inserts prosody; you´re saying it respects author intent
dw: lots of use cases for author decorative control
ia: further discussion tomorrow in APA meeting
mh: can demonstrate prototype
people in assessment community want a standard approach, support the ETS efforts
have some implementer interest
js: supports home invasive devices
mh: those are already ahead of this
we could offer them to import our kind of content
ia: beyond pronunciation, could help with language dialect issues
mh: have examples of improving chemistry pronunciation
(though chemistry navigation still needs to be addressed)
need to push implementers on using @lang info
gk: Japanese publishers really need this kind of improvement
<Zakim> MichaelC, you wanted to ask about assessment and implementer participation in TF
mc: what is prospect of assessment and implementer participation?
mh: that´s next now the TF is ratified
think we can get readaloud implementation; might need to push more on screen readers
trackbot, end meeting
This is scribe.perl Revision: 1.154 of Date: 2018/09/25 16:35:56 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00) Succeeded: s/coeeficient/binomial coefficient/ Succeeded: s/Peter/Dani/ Succeeded: s/nimmeth/Nemeth/ Succeeded: s/3 taxonimies/mechanisms to attach the metadata/ Succeeded: s/Micheal,/MichaelC:/ Succeeded: s/weel/well/ Succeeded: s/goog/good/ Succeeded: s/licencing/licensing/ Succeeded: s/DL/LD/ Succeeded: s/George: Json/Daniel: Json/ Succeeded: s/George: we do it/Daniel: we do it/ Succeeded: s/George: we are proposing/Daniel: we are proposing/ Succeeded: s/Daniel:/DanielWeck:/ Succeeded: s/George: have you approached to TTML group?/JF: have you approached to TTML group??/ Succeeded: s/gk:/mh:/ Succeeded: s/@@1/DanielWeck:/ Found embedded ScribeOptions: -final *** RESTARTING DUE TO EMBEDDED OPTIONS *** Default Present: janina, Joanmarie_Diggs, MichaelC, Irfan, Becka11y, IanPouncey, Roy, clapierre, pkra, arnog, mrobinson, MatthewBrennan, matt_king, mhakkinen, George_Kerscher, JF, STeve, lisaS, Gottfried, Marisa, Avneesh, Romain, Daniel, romain_, DanielWeck WARNING: Replacing previous Present list. (Old list: janina, Joanmarie_Diggs, MichaelC, Irfan, Becka11y, IanPouncey, Roy, clapierre, pkra, arnog, mrobinson, MatthewBrennan, matt_king, mhakkinen, George_Kerscher, JF, STeve, lisaS, Gottfried, Marisa, George, Avneesh, Romain, Daniel, romain_, DanielWeck) Use 'Present+ ... ' if you meant to add people without replacing the list, such as: <dbooth> Present+ janina, Joanmarie_Diggs, MichaelC, Irfan, Becka11y, IanPouncey, Roy, clapierre, pkra, arnog, mrobinson, MatthewBrennan, matt_king, mhakkinen, George_Kerscher, JF, STeve Present: janina Joanmarie_Diggs MichaelC Irfan Becka11y IanPouncey Roy clapierre pkra arnog mrobinson MatthewBrennan matt_king mhakkinen George_Kerscher JF STeve Found Scribe: Gottfried Inferring ScribeNick: Gottfried Found Scribe: STeve Inferring ScribeNick: STeve Found ScribeNick: clapierre Found Scribe: Charles Found ScribeNick: romain Found ScribeNick: clapierre Found Scribe: Irfan Inferring ScribeNick: irfan Found Scribe: MichaelC Inferring ScribeNick: MichaelC Scribes: Gottfried, STeve, Charles, Irfan, MichaelC ScribeNicks: Gottfried, STeve, clapierre, romain, irfan, MichaelC Found Date: 22 Oct 2018 People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option. WARNING: IRC log location not specified! (You can ignore this warning if you do not want the generated minutes to contain a link to the original IRC log.)[End of scribe.perl diagnostic output]