Minutes: MathML Semantics meeting 13 May

Meeting was recorded:
https://benetech.zoom.us/rec/play/v8Aoc-qs_DI3GNOSsQSDAfUoW43sfams0SUZqPcKnkrkUnQAYFL0YbESare2vGxc5xbKTv1tsqsSqtBv?continueMode=true&_x_zm_rtaid=v_RHcOFeR6ejJLMYlrqMWw.1589396106010.488fa735cf75888459df382262277346&_x_zm_rhtaid=451
(Access Password: 7D.2hC+8)

Attendees:

Neil Soiffer

David Carlisle

Sam Dooley

David Farmer

Bruce Miller

Charles LaPierre

Patrick Ion

Murray Sargent

Akashdeep Bansal

Thanks to Patrick Ion for taking some of the minutes.
Feedback from Chemistry CG

NS: There is a difference between equations and formulas.  Formulas can
occur in equations.

= may mean “double bond” in formulas.  Thus, subjects need to go on mrows
(and not just on the math tag).

NS: Do they need different subareas?  (organic, for example)

NS: Chem spreadsheet with examples:
https://docs.google.com/spreadsheets/d/1h-8k_bwQ1bO7gusb0O2DfcEs0DZUAo6QVGYZuVjWVB8/edit#gid=0

PI: Do they use ChemML?

NS: ChemML supports arbitrary data,and is complex, so publishers are
unlikely to use it. Chem Formulas use math notation. People do this with
TeX and WYSIWG editors.

PI: So this is a convenience markup?  Like math at the simpler level.

NS:  nuclear decay, chemical bonds, all can be done in MathML.

NS: in chemistry, lower case and upper case letters are spoken differently
(in a special way that is different than some math screen readers).

NS: I was a little surprised they want lower case/upper case spoken
distinctly despite screen readers having a way to speak their difference,
but they want AT to be more explicit case distinction for chemistry

PI: There was a stuttering convention for case distinction…

NS: Anyone want to object to allowing ‘subject’ elsewhere beside <math>?

[no one]

PI: we need examples---as we do for math

NS: we await more information from them.

DF: the “=” for a double bond could have been marked up with a mathrole

NS: yes, but in a formula, it is always a double bond. Could try to unify
formula and equation but that puts extra work on the software/user to
distinguish the bonds.

DF: nobody is writing the MathML. Someone is using a tool that is writing
it for them.

NS: Agree. Software could do transfer more than the subject area. Not done
at the moment.

DC: Disagree. Our technical editors at NAG do it and they get context
special help to write it which they don’t get with TeX.

MS: we have thought about allowing chemistry in math modes. One thing is
that you don’t italicize variables in chemistry formulas.

NS: Good point. The semantics affect the display. You could do that easily
with CSS, so doesn’t strongly affect the implementation.

SD: For mathrole, if they all started with “chem-”, we could catch all of
them

PI: there are coefficients in front of the formulas, and they need to be in
italics.

NS: good point -- I think CSS can do it.

PI: again, examples would be good

MS: you could use the math alphanumerics...

List of potential subject areas (along with whether these go into the spec
or a note or elsewhere)

DF: here’s some subject areas [went through some briefly…]

Preliminary list of topics, and some examples of how the topic can
disambiguate meaning.  The statement that x "means" something, there is an
implied "usually".

sets (in the sense of basic set theory children learn)

  \cup means union

  \cap means intersection

  | or : occur in set builder notation

algebra (in the sense of what children learn after arithmetic)

    \times means multiplication

  equations (as in: solving them)

  graphing

      (a,b) means a point in the x-y plane

  inequalities

      (a,b) means the open interval a < x < b

      |a| means absolute value

trigonometry

   f^{-1} means inverse function

calculus

  onevariable

     | occurs by itself to mean evaluated at

     \times means multiplication

  severalvariables

     \times means cross product

differentialequations (ordinary)

pdes

linearalgebra

    \times means by (as in 3 \times 5 matrix)

    A^t or A^T means transpose

analysis

    |a| means absolute value

  realanalysis

  complexanalysis

    x^* or \overline{x} means conjugate

combinatorics

   \times could mean multiplication, or Cartesian product, or "by" as in 3
by 5 grid

          (so guessing is wrong more than half the time)

   (a,b) means an ordered pair

   |a| means cardinality

probability

   P(a | b) refers to conditional probability

NS: I would like to minimize the number of subject areas as that would make
life simpler.

NS: For example, at least from the example, I don’t think “trig” needs to
be one because sin^-1 is always an inverse function in every subject area.

SD: We should think about “readers”, not producers

DF: you need to get to higher math above calculus to get to some of the
conflicts.

BM: We might be doing this backwards. The subject areas choose the
defaults, but we should think about actual values that need to be
specified. AT versus translating to a language. What about reading in other
languages.

BM: I’m unclear about whether this is just for reading it in English. Does
it need to work in other languages too? For computation, just for
Mathematica or other systems. What’s the overall goal?

NS: For me, the answer is “yes” to all.

NS: If we wanted to specify just a reading, text doesn’t work because the
needs of someone who is blind is different from what a person with dyslexia
needs. “start/end fraction” is need for someone who is blind, but bad for
someone with dyslexia.

NS: I want something that helps translating to mathematica or maple. It
should say how to do the translation, just a clue that if you translate it,
here’s something to tell you what it means to aid in your translations

BM: I’m not saying putting how the words are composed, just that ‘t’ is
time or ‘t’ is transpose.

PI: Indeed, whether saying “transpose of a” or “a transpose”

NS: Yes, I want to disambiguate so I don’t just say “t”. Or that a
computational system that knows to translate to their transpose function.

BM: I think we already have a system for that: content mathml. At least it
is used more than whatever we come up since it exists.

NS: But parallel markup never flew. We need a better solution that is
simple for people to add if they want.

BM: the difficulty is to say what the math means so you can compute with.

BM: that difficulty exists no matter the representation.

NS: I agree mathrole is likely to be able to mapped to another format.

SD: I want to round-trip ContentMML <-> PresMML

NS: CMML hasn’t caught on we have to agree; allow hints on PMML to make
things better now.

This is a different job to CMML’s; a lesser task.

BM: How would this new form of Content Markup improve things.

NS: For the first 5 years, say, a publisher might disambiguate symbols by
adding the subject area.  That’s easy for some to do.  Problem for
accessibility and translations is you don’t have the context of the
document.  It’s hard to look upwards usually, though it can be done in
principle.  Adding to every math expression would be done with simple regex
handling.

Bruce, you’re still sceptical?

BM:  A lot of amazing machinery might be needed to analyze enough for every
symbol.  In MathPlayer i have done mode changing (e.g. to chem for speaking
things differently).

NS: perhaps tagging or tooltips showing properties of Hydrogen.  Perhaps in
calculus one

Might know it was a derivative.

BM; accessibility versus computability; latter seems unrealistic

DC: theorem provers have large proof terms that are difficult to render ;
this might

Allow hints that would help them to read back their proof terms ina
semantic way

BM: it's system specific ---

DC: it doesn’t have to be; but a terabyte proof-term has other problems

NS: the more typical case might be a bit from WikiPedia that someone wants
to put in

A CA system.  Hints can allow a lot more to get done or correctly
translated.

NS: you’ve been working with SRE, Akashdeep?  Would this help?

AB: you can define the domain there; I can check with Volker as to what he
allows

as values.  That might help authors to be able to specify subject.

NS: I thought it might be on a <math> element  initially

Could you find out, Akashdeep, what subject areas have been found to be
useful? DF’s analysis begins from books.

AB;

NS: we have  divide; Maybe the explainer would be useful for Bruce, say.
AT the start of the call

I was talking to Sam about what should be in the AT; there’s a new W3C
group on the Accessibility Platform Architecture; I’m joining that group,
as is Charles.   Having documented examples of what we need will be very
good.  I’ve co-opted Sam to help write this.  Any others who wish to
contribute should pipe up.

SD: We’re going to document what we think has to be transferred for math
into the Accessibility Tree for example purposes.  We need to explain why
it’s not all simply text.  Perhaps they’ll see

extra ways to get some of their own problems solved.

AB: Math is not simply all text at least because long expressions
necessitate some navigational help.

NS: Yes, but those not familiar with math won’t immediate;y think of even
this.

DF: Will the chemistry be in this doc

NS: Yes, a good idea

DF: Chemistry (or much of it) can be seen as the same problem

SD: Yes

Writing an "explainer" about math and accessibility

PI: I have a question that goes back to Bruce. What goes into the
accessibility. I couldn’t find anything and now I know why -- they haven’t
defined it. I gather that there’s a group to specify that. So it’s ideal
that the Math group be involved. There are three things: MathJax should
understand what it should do/render; there’s the dream of computer algebra
systems of being able to use it, and there’s the accessibility issues. I
didn’t even know about the different accessibility needs.

NS: that’s why we need the document to inform everyone.

PI: I discussed with Bruce the various options, but what shows up in the
accessibility tree affects what we do with microdata or mathrole.
Expressed as JSON-LD or whatever; but these are technical puzzles that must
come later.

CL: I’ll help out but don’t have a lot of time. We also need to be aware of
the “Read Aloud” functions of EPUB for reading MathML in reading systems
(currently they either skip the math, read just the text no semantics
(fractions etc.), or speaks the alt-text description inside the mathML.

NS: we need to come up with something that explains what’s needed for math
accessibility for a group of people who don’t know about math accessibility.

NS: I’m volunteering SD to help.

SD: agreed. Where shall we do this.

NS: use google doc or wiki (https://github.com/mathml-refresh/mathml/wiki)?

SD: Google docs are easy

NS: let me see if the wiki supports math.

NS: We also have our main page with links: https://mathml-refresh.github.io/