Minutes: MathML Semantics meeting 28 April, 2020

Meeting was recorded:
https://benetech.zoom.us/rec/share/1chTMZK31nNIZrPL8WbkZL88EYnPaaa81SgeqKcIy0nEu2KW_pm_5JTbSWIbv0M5
(Access Password: 2H!4$N2@)


Attendees:

Neil Soiffer

David Carlisle

Sam Dooley

Akashdeep Bansal

David Farmer

Murray Sargent

Bruce Miller

Charles LaPierre

Patrick Ion

Thanks to Charles for taking most of the minutes!

Subject Area breakout.

(highschool / undergrad)

Need a list - David F. can help

NS: Brian Kardel felt we should look at other ways to add semantics,
already in use in Web.

NS: can’t use aria role because it needs to map to A11Y Tree and nothing to
map to other than “math”, which is useless.

NS: Brian suggested thinking about Micro Data / Formats, JSON-LD, schema.org


Micro format use classes, (person - record) etc.

PI: Maybe use MSC and link to (Math Subject Classification) (values of a
subject area) 6000 leaves, but too fine grained.  It is a tree so you can
scale it appropriately. (62 general subject areas).

PI: MSC http://msc2010.org for MSC2010 now revised to MSC2020 at
https://mathscinet.ams.org <http://mathscinet.ams.org>/msnhtml/msc2020.pdf
and online at MathSciNet https://mathscinet.ams.org
<http://mathscinet.ams.org> and zbMath https://zbMATH.org
<https://zbmath.org> (soon to be free as of Jan 1, 2021)

There was a Linked Open Data version of MSC2010.

NS: Micro data pointing to schema, do we have enough structure to make
sense?

PI: National Anthology group in Buffalo has a list that may work.

PI: OpenMath would be too complex.

DC: Microdata in HTML5 no scoping to refer to OpenMath and would need to
put a Full URL every time you use it for every element in every MI, which
no one will do (unless their input system does it for them).

NS: Want one set, so that we don’t have to figure out what it is pointing
to.

NS: But notes can say “this is the same as XXXX in OpenMath”.

DC: On microdata names:

https://html.spec.whatwg.org/multipage/microdata.html#selecting-names-when-defining-vocabularies

CL: sorry need to jump off.

NS: For AT you need to know what it is so that it can be pronounced
correctly.

NS: \times could be “times” or “cross” or “by”.

NS: Dozens of contexts.  Thousands of specific interpretations. (1100
heuristics in Math Player,

Plus ways to guess the context.)

PI: Vertical bar is a good example:  it will need the math-role often.

(Infix, fence, …)  It is an example of something which does not have a
useful default.

Sam: I would like it to be possible to put a math-role on everything.

Vertical bar may not have a default meaning.

MS: Don't use vertical bar for “norm”.  Use the norm character.

INS: if you don’t put a math-role, need to have a way to indicate that the
default is intended. Should there be some attr on <math> to say markup is
right and don’t run heuristics?

DF: What is |A| means “cardinality”.  That could possibly be the default if
the subject was declared to be “combinatorics”.

BM: if math-role is used somewhere in the document, maybe take that as a
signal that guessing is not needed.

SD: the AT decides whether it trusts the author or not.

BM: Must be able to express wrong mathematics!

?: Math-roles come from some finite list. But what happens when we
encounter an unknown attribute?

?: That is another argument for all using one selected list.

PI: Is math-role single value or multiple values?  E.g., “cardinality
msc:05C” Maybe not a good example.

(or maybe OpenMath: or DLMF: colon)

NS: Back to our start:  what other options are there?

PI: JSON-LD is based on context:  can point to an ontology and give the
context.

OpenMath has an ontological aspect.  Need something to point to.

BM: don’t lock down the set too much.  A URL does not mean you need to
follow it.

The Marketplace can decide which sets of terms are useful.

DC:  that is why you need namespaces.

PI: We don’t want to write our own ontology.

NS: Informal definitions are adequate:  we provide names and what they are
intended to be used for, categorized in groups.

[discussion about multiple values in math-role]

[some discussion of preserving the authoring system semantics. Eg.,
Mathematica and using semantics tag to do that vs using mathrole].

DC: the problem with <semantics> around the entire math is that once
editing begins, it is meaningless. Mathrole is more likely to be kept and
still be meaningful.

NS: We will meet again next week. Please send me agenda items if you feel
we are missing something.