Minutes: MathML Full meeting 25 May, 2023

Attendees:

* Neil Soiffer

* Louis Maher

* David Carlisle

* David Farmer

* Sam Dooley

* Bruce Miller

* Patrick Ion

* Bert Bos

* Deyan Ginev

* Paul Libbrecht

* Cary Supalo



*Agenda*


* 1. Announcements/Updates/Progress reports*

NS: Due to vacations, the MathML intent group will next meet on Thursday,
June 22, 2023.



DG: Do we have a charter extension for June?



BB: The current charter will be extended while the new charter is being
renewed.



NS: Do we have a new Zoom link we should be using?



BB: Our current MIT link will work for a couple of months. BB will give us
a new W3C Zoom link which NS will put on a new calendar entry.



**ACTION** NS will send out the May 18, 2023, intent meeting minutes.



NS was wondering if SN had made progress on the MathML current state
description paper that SN is writing for xml.com.



DG: There's a new browser that supports MathML core this week. It's Samsung
Internet. They hold just under 5% of the mobile browser market share.  So,
5% of mobile users use Samsung Internet on Samsung devices. And this week,
version 21 brings along Chromium version 110 Which has, the maximum core
support. I tested it and it works great.



DC: Do you know if they put a math font on the device or whether we use web
fonts?



DG: It does not have math fonts.



*2. Some progress on ":structure" -- feedback?*



NS did some work on MathCAT on the intent structure property. The intent
structure property turns off all the assumptions that the AT would normally
make. Only the structure of an expression is described. NS has not fully
implemented the structure property.



DC: had added a column to his "Intent Examples" document showing the effect
of the structure property (https://w3c.github.io/mathml-docs/intent-examples
).



DC's columns are: MathML, MathML (no intent), MathML (structure), MathML
(with intent), TeX, and Comments.



DC: Consider X transpose. If you don't do anything, it says X transpose.
Whereas if you put structure, you get X super T. Or then with intent you
can make it say various things like transparency or trans.



NS saw a couple of mistakes made by MathCAT. MathCAT generated the
intent=:structure column.



NS: was looking for comments.



DG: The F-star structure example shouldn't have an audible superscript
pronunciation because you don't know what that superscript is in the
structure.



NS: There were a few things like f hat, f star, f prime that I did not do F
superstar, F super prime, etc.



NS: So, this is when you know what the star is. The thing you have to
contrast to is you have a lot of examples where the star is used on the
baseline as an infix operator.



DG: You don't know whether the stars are on the baseline unless you say
something for the superscript.



The group reviewed DC's table and gave NS comments about it.



PL: The structure property gives you an alternative pronunciation that is
closer to the reading.



PL: Perhaps a JavaScript could be used to allow the reader to turn the
structure property off or on with each equation. This would be a button in
MathCAT to do this.



NS: it might be a hassle to read out an equation both with and without the
structure property.



PI: What do you want the AT to do if you turned things off.



LM: You would hold out for the structure which would give you a good chance
to understand the equation.



LM: The structure property is turned on and off by the author and not the
listener.



LM: All you want to do is copy the exact structure and figure it out later.
Because if you have to do it orally, if you can't see it, then you got one
shot to get it right and you want the structure to be told to you.



LM: Now if you can see the thing or go back to it at your leisure, that's a
different story. I was brought up in a classroom where there was no written
material to explain class lectures.



Dg: It's good that there's a fraction here. I wanted to bring up the
question of attributes for the structured or readouts.



DG: So, if a fraction has a different line thickness than the default may
be different than if the line thickness was zero.



NS: MathCAT does not readout line thicknesses.



DG: Would we expect the structure property to alert the listener to
attributes that are of specific significance?



NS: Well, so that's a good question given that the whole point of the
attribute is to Change what it looks like.



NS: Well, the other one I was going to mention is if someone does left or
right spacing, should it be spoken?  NS thought not.



CS: If you put sodium in an expression, is it going to read it as "Na" or
"N" "a". NS: it is the second.



DC: We might change the default behavior in the case of chemistry. If we
see "Na" we probably would want the AT to say it as "N" "a" instead of
pronouncing "na".



DC: We need for the author to turn on the chemical default's pronunciations.



NS: We have not defined either common or legacy responses.



DC: If you specify conflicting properties such as common and structure,
this is an error, and you get what you deserve.



DC: When you speak an expression, intent wins.



NS: wants to move on to consider legacy and common.



DC: is worried about legacy.



NS: Legacy has to be the default.



DC: There was no legacy cross-browser to break.



NS: JAWS speaks math differently than MathCAT. Vispero is writing a JAWS
style of pronunciation for math.



BM: does not know what the public is expecting from AT tools, but he is
uncomfortable about what legacy means without MathML. You would need to
know what type of material.  He is unhappy about MathCAT guesses.



DC: Says there is no default behavior that works for all browsers.



NS: MathML 4 gives the author the way to specify how things are to be read.



NS: Most people who read math are not math experts.



DF: It is rare that the listener wants to hear vertical line "A" vertical
line. The listener should get to choose what they hear, especially if the
author did not put any intent. He is OK with mathcat to read things like
they always have been read.



DF: Our loyalty is to the listener and not the author. The listener should
be able to choose what the listener wants to hear, especially if the author
did not give any intent.



DF: Reading symbol by symbol is a rare case. People would not like it as a
default. The reader should be in control.



NS: You should not require the AT to have many ways of pronouncing things



PL: I agree that, for the authors, it makes sense to decide for the
listener, when something should read the structure.



PL: Also, we should have keyboard shortcuts telling the AT when to read the
structure and controlling reading navigation.



DG: Has there ever been a discussion about versioning MathML say by using a
version attribute? HTML has a way to specify the version of the spec. The
AT would look at the versioning attribute.



DC: HTML versioning was a problem. The browsers never implemented
versioning.



DG: If there was a clear way to indicate this is MathML 4, then we could
set the defaults. Everything that was lower than MathML 4 would be treated
as a legacy expression.



DC: What would you do if you had different MathML versions in the same
document.



NS: If you had common and structure as the only two modes, what would be
the default?



DG: Structure.



DC: Common.



DG: If you turn on common on archive, it will be a complete nightmare,
structure is at least 10 times better.



DG: It depends on the material you're listening to.



DC: It is a good thing to be able to turn off the defaults, but I don't
think What you actually want is a different default like chemistry or even
archive or some other default.



NS: He said that some blind people told him that it was alright to say
square because the blind person could figure it out. That is a simple
superscript case.



NS: I think if it says absolute value and it's really something else, this
would be a problem.



NS: Any resolution on legacy, common, and structure, and what the defaults
should be?



DC: Add them anyway and we will continue to talk about them.



NS: We have not defined what common is.



PL: I definitely support the default to be common in our structure for
non-identified content.



BM: Structure reading can be painful.



NS: In common, vertical bars would be ambiguous, whereas superscripts would
be defined as power.



BM: A default seemingly should be as concise and natural as possible
without assuming any semantics.



NS: I think you now just defined a homework assignment for yourself for the
next 3 weeks is to write that down.



DC: I think we need to start by writing the common down.



DC: We need language to specify what these defaults are.



NS: I wrote a proposal for what should be in common a while ago. It is
here: (
https://github.com/w3c/mathml-docs/blob/main/minimal-intent-core.md#common).



NS: There is also an issue where we started to discuss this: (
https://github.com/w3c/mathml/issues/433#issuecomment-1420212046). That's a
good place to continue the discussion and refine my proposal.



NS: If common is defined in every case, then the content generator knows
when he must add intent to override the common pronunciation.



BM: MathCAT is impressive. The second column is a good example of what
common should be.