[public-mathml4] <none>

Agenda:
https://github.com/mathml-refresh/mathml/issues/91#issuecomment-560946753

Meeting was recorded:
https://benetech.zoom.us/recording/share/7dOV6ZeUQIKvQy33r_pQBnr0qz-GFV3euZu_Usl5-gSwIumekTziMw

Attendees:

Neil Soiffer

David Farmer

Sam Dooley

Steve Noble

George Kerscher

Avneesh Singh

Jason White

Charles LaPierre

Murray Sargent

Regrets: Patrick Ion

Thanks to Charles for taking notes!
Questions from George Kerscher

   -

   Chemistry semantics and what support does the MathML CG need from the
   Chemistry CG?

Neil: document based on subject area.  Need feedback from Chemistry
Experts.

Sam: vocabulary of chemistry constructs, having such a list would be
valuable.

Neil: these are google docs /sheets

Jason: have issues with comments, but I typically export it. I can make
comments.

Neil: we will circulate this with minutes for all the links.

George: we have a chemistry call and I can link to the Google docs/sheets
that you want us to review and get the chem group.

Avneesh we did a doodle poll 12 or 19th we don’t have the results yet.

George: we can get the review that you need.

   -

   What is the expected timeline for MathML support in Chromium? This
   relates to the plans to move the MathML CG to a formal W3C Working Group.
   The Igalia fork of Chromium with the current implementation is available
   now, but when would this begin to be included in nightly builds of Chrome
   proper?

Neil: We have discussed this in the core meeting. It is an unknown. Maybe
fall of next year, but it could the summer or even the following year.
Reviews are being done quickly by Google, which is a very good sign to me.
No promise that it will go into chrome though; same for Microsoft not a
give-in.

Charles: MS is interested but will not say publicly they will implement
what Igalia’s MathML implementation but they are very interested in it and
hopeful this will be their solution.

George: we do have some influence from the accessibility side of Microsoft.

   -

   The publishers are using MathML right now and it is unclear how this is
   working in Reading Systems and with screen readers.

    Neil: I don’t know how its doing in EPUB RS. I had to help my son with
a calculus proof, and I needed to review some math to help him. I was
surprised that maybe 80% of the pages I landed on were accessible using
NVDA and mathplayer.  Only few places were PDFs or images. So math
accessibility has taken a big leap in the last few years.

    George was this VS?

    Neil: Wikipedia, random sites, stack exchange, blog posts etc. all
worked with NVDA. Not sure how the mathML was implemented but mathJax
rendered most of the pages. It hides the MathML but exposes it to AT.

    Avneesh: MathJax V3 has the default off. V3 has a workaround to turn
this back on it is buried in the menu system

    George: Charles and I had a meeting with Volker, we showed him word to
EPUB for MathML and Volker was happy with it.

George: I showed Neil some output from Macmillian which had problems
reading. What we think is happening is that we are being read the
aria-label, but the code looks good with proper namespaces for the MathML.

    Neil: I think the aria-label takes precedence.  Oh thats right they
have role=”math” which breaks how NVDA handles the math.

    Avneesh: this is very good information and we need to write some some
best practices.

    Jason: sounds like a NVDA bug report

    Neil: Sina, Charles, Jason, and I came up with an accessible solution
that uses some JS. The DIAGRAM center on this, why can’t publishers use
that?

    George: we have that in our test book. What about adding alt text on
the math element?

    Neil: mathML has always had alt-text and alt-image and no one has ever
used them.  Not sure why JAWS wouldn’t pick up the alt text on the MathML
before it finally supported mathml for real.  I suspect the spec will
continue to support these two, but they wouldn’t probably be required.

    George: we have influence on publishers and Benetech is doing
certification and we can create a best practice, and we have this
WordToEPUB which is being used and once we have it signed we will make a
press release, and it puts the MathML in the exported document and the
LaTex in the annotation.

    Neil: that's great that it is in the annotation and not in the alt-text

    George: we would also include the alt-text filled out as well.  I see
Macmillan / Pearson, and VitalSource (VS) is supporting MathML with
MathJax.  Thorium is using MathJax 3. Seems really complicated and I can’t
figure out what is going on.  We need to get it worrying correctly.

    Avneesh: alt-text is an alternative, RS/SR that reads out the MathML
and repeats the alt-text

    Neil: if someone understands the math Ml otherwise the fallback would
use the alt-text

    George: Read-aloud could grab the alt-text and speak that while that
MathML is being displayed.

    Neil: that wouldn’t be ideal, for sync highlighting to know where
everything is.  You could do the whole blob. Mathshare by Benetech is doing
the sync highlighting with MathML using Volkers SRE.

Neil: Good point what happens with alt-text and MathML if you do understand
the MathML and when you don't.  Must still get the Reading systems who
don't understand MathML to grab and present the alt-text.

    Steve: publisher created content / workflows: I assume mathFlow is
being still used in some houses (not sure name) has that ever incorporated
Volkers SRE?

    Neil: MathFlow uses a version of mathPlayer to generate alt-text

    Steve: on publisher side workflow / products like mathFlow to create
all things for Screen readers and TTS that we could provide that input.

    Neil: mathrole or subject area, allow subject area, they could specify
that in a command line in their workflow.  To get mathrole in the content
may be hard to do. That would be MathType / Wiris. Once you got the final
document you could hand edit those equations which don’t speak well to add
these roles manually.  Authors should do it initially.

    Jason: general mathematical for the entire document vs. single
equations.

Neil: I think publishers who care; will add some extra mathroles if the
tools doesn’t add them automatically.

Jason: authoring tools / checking tools and different output rules could
request info from the author.


David: google doc
<https://docs.google.com/document/d/1cZnff5_fi_ucNyZ1ex2msmJLE55FAZD-QInkLYe8xiE>
about
PreTeX

David: not a lot of work for unambiguous LaTeX  binomial coefficient should
have a macro 2 args better to use that macro they are already writing
semantic LaTeX.  Author doesn’t have to do a lot to be semantic. π(x) is a
prime counting function.  Pi as a function vs. π * (x) .  Not a burden for
the author.

PreText authoring system (XML) human readable/writable captures structure
of documents.  100 books in preText. Open source books. Looking at what
they write vs. if they are semantically rich.  Written semantically vs.
what is preserved in the book afterwards.

I can convert what the author wrote to a semantically rich version.

I am getting close to understanding how to go from PreText to semantically
rich MathML.

(a,b) - could mean open interval ab, ab in the Cartesian pair, ordered
pair, …

This could occurs interchangeable in the same book ⅓ of the time means
different thing.

David:  Examples of books in PreTeXt <https://pretextbook.org/examples.html>

Neil: somehow convert those macros the semantics meant by that macro into
what we are proposing as a mathrole. We are not sure exactly where, may be
a leaf, could be on an mrow, etc.  where that mathrole goes. It is not a
fixed point and we need to figure out the details. David is looking at real
textbooks and real data, and that is very valuable and what the defaults
should be based on real textbooks.

Jason: some of the TeX to MathML and I am not sure which is being used
right now.  2 kinds (processes the source LaTexML written in pearl (expands
macros etc.) tex process includes metadata which could extract mathML /
html could generate the desired output.

Neil: Bruce Miller is part of the group and could help us in this area.

Jason: I have used LaTeXML I like it.  I could test that

Neil: we need a spec first.

David: this PreText can be converted into MathML.

Neil: need to hook you up with Bruce and see how PreText works with his
tool and get Volker if he is interested in using his macros vs. something
more standard.

David: I need to make a big table which has the semantic meanings etc in a
useful form.

Neil: google doc
<https://docs.google.com/document/d/1NrPFIgxt6FNAOoWzSunI71auADfssCGkiAuDso7RI6s/edit?usp=sharing>

Neil: Chemistry is in the doc near the bottom. I tried to write out
notations that are more specific to a subject area.  E.g., in calculus and
you see an integral sign, you know what it is and you probably know that
without knowing the subject area.

Neil: Something that I think we need to discuss is how much of a pattern is
needed before you need to include the mathrole?  Right arrow that appears
randomly in an equation vs. something that is expected in a given pattern.
E.g, if the base is “lim” do you need a mathrole to tell you that the right
arrow is “approaches”?  Chemistry is an area where it gets a little fuzzy
ie. H + Cl. I wouldn’t know what to do until you see H + Cl -> HCl, then
you understand this is chemistry, so what is the line between obvious vs.
open for interpretation.

Neil: please add to this document if there are areas missing.

David: I think its a design flaw in TeX that you have to put a /, before a
Dx, I would rather see just a space before the Dx

Neil: you are worried not only about the semantics but also the display.

Murray: Neil got this into unicode and the differential D /dd which is
unicode 2146 which automatically puts the /, when needed.

Neil: Normally ambiguous are the invisible times , invisible + like 3 ½
some which doesn’t require a mathrole.

Jason: we may have to specify a practical limit when a role should be
required.

Neil: looking at the entire document may be too ambitious , Volker is
looking into this.  I would suspect that looking at a mrow, fraction
construct that might be the limit. There are probably some examples that we
could infer meaning.  Ie. if you use a $ then you know that the # following
is a monetary, but you would say “3 dollars” vs. that may be just a speech
requirement.

Neil: interpreting the semantics is hard, and it won’t get easier, rules to
try to figure out the semantics. But if authors do this will help this
process.

Next meeting in two weeks.