Minutes: MathML general/semantic meeting Dec 12

Agenda:
https://github.com/mathml-refresh/mathml/issues/91#issuecomment-566272794

Meeting was recorded:
https://benetech.zoom.us/recording/share/ku69jWIPrFxZKnlyDtsl-6A8Tx5d-jyv1C94WWFH5KOwIumekTziMw
Attendees:

Neil Soiffer

David Farmer

David Carlise

Bruce Miller

Sam Dooley

Steve Noble

George Kerscher

Charles LaPierre

Patrick Ion

Murray Sargent


George: want to make sure Chemistry group has requirements, may start a
google sheet.

Neil: would like an example of how Physical Chemistry might conflict with
General Chemistry.

George:  want a spreadsheet with pronunciation information.

Neil:  Ambiguities to resolve:  conflict with math (e.g., arrows), and
conflict between areas of chemistry (maybe such conflicts do not exist?).

Patrick: asks about relationship  with cml,

Neil: target is mathml describing reactions or just molecules within cml
not the full structure encoded in cml

Neil: Update from Sam

Sam, no updates looking to make decisions about markup, to use (exact list
of names can be tweaked later)

Neil: Updates from DavidF

David F: added two pages to my document
<https://docs.google.com/document/d/1cZnff5_fi_ucNyZ1ex2msmJLE55FAZD-QInkLYe8xiE/edit>
.

(In combination with Bruce) we can produce latex macros to encode semantic
information but do not see how to map to roles in mathml markup?

David F: If the latex source tells you that (a,b) is an interval and later
it is a cartesian point.

David F where/how do you look up the mathml role?

Neil: we need to produce (in the spec or elsewhere)  a document that lists
the currently defined roles. Also we need to produce one (shorter list) for
subject areas. How you go from tex to the math role is down to the author
of the conversion software.

David F: hope someone makes up a document showing examples of markup

David F not that many ways to make mathml (I only use mathjax)

David C: there needs to be support in the convertor and it’s not likely to
be highest of Mathjax priorities

Neil: Bruce’s convertor may be more likely to support this in the timeframe.

Bruce: converting from tex to mathml isn’t trivial

Neil: two possible scenarios...David F, if your macros included the name of
the mathml role to be used, then the latexml converter  could add the role
to the generated mathml, other way latexml does more work and maps the tex
names to mathml roles. First is possibly more realistic.

Bruce: yes that is the right model, need to get the infrastructure to get
the math roles to be copied from tex to mathml, this is similar to dlmf
markup where we have several hundred special function macros, had a
function-defining macro that had several keywords, one of which could be
mathrole which would be ignored in tex typesetting but could be used by
latexml.

Simple approach would be for DavidF to have a pretex style file with a new
definition command with a role keyword.

DavidF would need to see more examples.

DavidF if we could do abs value, and open interval and function
application, it would be a lot clearer.

Neil: can you make up a small document including a small set of tex
examples and converting to (arbitrary) math roles, just to test to workflow.


DavidC I could probably add roles to the NAG documentation as an example.

Bruce need to clarify where the role attribute should go.


Neil: if you have a transpose do you put mathrole=transpose on the msup or
the mi T

Patrick: norms (valuations) often have subscripts ||A||_2 ; intervals can
have them as well

Neil: Steve Noble added some Chemistry to my document, I added information
about what to do where there is a notation that does not have a mathrole,
but you want a specific meaning or spoken form

Neil: we could add two other things ‘alt=text’ allows spoken text for that
construct. (eg on a token or mrow) This would over-ride anything inferred
from the structure of the content. Probably would not use this for braille
since braille is syntactic and one doesn’t need to provide a meaning/way of
brailling it.

Neil: secondly, for computation, authors may have matlab (or nag) in mind
and could tag  the construct with an expression that can be evaluated,
data-nag=... data-matlab=...

DavidC this has a big overlap with the semantics attribute.

Patrick: alt (speech) could need multiple languages

Neil: normally the author has a language in mind.

Sam: I prefer this to <semantics>, avoiding the xml overhead

Bruce: probably unrealistic to annotate presentation to be able to compute,
annotating for accessibility and spoken forms is more realistic.

Patrick: it’s your math education that makes you an inferring converter so
as to be able to read math papers

Neil: last topic: Names for the math roles

We agreed on mathrole, but not said much about the names of role values. We
want consistency of naming. Is OpenMath a good source of names?


DavidC: openmath symbols  https://www.openmath.org/symbols/

DavidC: Openmath is by design very open ended and the CD system allows
people to name operations without reference to other CDs so may be used as
a source of possible names, but not as a fixed reference for a set of names.

DavidC: Also, the names are namespaced, so one dictionary doesn’t really
care about the other dictionaries, so OpenMath names are not necessarily
consistent across dictionaries.


Neil: In New Year we should decide on a scheme and then start naming
specific roles.

Patrick: Can you include your geometry example here for me to think about:

Neil:

<math subject=”geometry” xmlns="http://www.w3.org/1998/Math/MathML">

   <mover>

      <mrow><mi>A</mi><mi>B</mi></mrow>

          <mo>&#xAF;</mo>

   </mover>

   <mo>&#x2225;</mo>

   <mover>

      <mrow><mi>C</mi><mi>D</mi></mrow>

      <mo>&#xAF;</mo>

   </mover>

</math>



Next Meeting: Jan 14th