Minutes: MathML general/semantics meeting Oct 8, 2019

We had (IMHO) a very interesting meeting where we got into the details of
semantics thanks to the contributions of Sam Dooley and David Farmer (see
links in agenda). We will continue the discussion next week.Meeting was
recorded:
https://benetech.zoom.us/recording/share/Ykkm5O8SLj3CruUy6ATJHN1a-91LHhg8u2mDj1Eq8g2wIumekTziMw
Attendees:

Neil Soiffer

Steve Noble

David Farmer

Charles LaPierre

Sam Dooley

David Carlisle

Regrets: Murray Sargent


Agenda:
https://github.com/mathml-refresh/mathml/issues/91#issuecomment-539119949
Discuss #141 <https://github.com/mathml-refresh/mathml/issues/141>

SD: I wrote a massive spreadsheet with the interesting stuff at the end of
the spreadsheet.

SD: Want a way to make sure semantics is part of presentation MathML.
Content MathML hasn’t been used much.

SD: Use ‘role’ to attach a role to every operator to bring a meaning to
MathML.

SD: The symbols I added came from an editor that I wrote.

NS: You added meaning to non operators also.

SD: Apologies that things should be under emphasized take up the most space…

SD: They are there so they can be referred to. I meant every node in the
tree can have an id.

SD: Allows one to switch from presentation to content mathml easily.

SD: Also works for DF’s PreTeXt markup.

DF: I want to preserve all the info in the PreTeXt

SD: It should if we agree on what all the names are

SD: Doesn’t prevent a semantic inference engine from creating or augmenting
the info in the MathML

NS: My problem is that you put too much info in the tree. In ARIA you only
add info when it isn’t clear from the tag.

SD:

NS: You don’t need to say an alpha is an alpha

SD: There will be a default: pi is \pi unless you specify otherwise.

DC: In the spreadsheet. It has something for role=norm, I get it. But what
about bold face “a” using the unicode ‘a’ for it. Why is there an id for it?

SD: It simplifies semantic inference

DC: but it can only be a bold a.

SD: there is a tension between how much we specify vs how much we assume

DC: what if I added role ‘boldb’ on a bold a?

SD: someone would do that for some real reason…

SD: There are several integrals. You can infer them, but I am trying to
remove all inference

DF: I have a question about bold R. How do I indicate that it is the real
numbers

SD: You add role=reals

SD: The id column is what the role value would be

SD: this is just a first pass through the symbols on the editor I wrote

NS: I want to get back to why so many roles. It make sense for Reals. What
about numbers? You don’t add a role for every number. So you do do some
inference.

SD: I agree there is an open question of how much is explicit vs inferred

DF: Can there be multiple values for a role?

SD: I haven’t done that

DF: if my generic bold face “R” is a ring, then I want to indicate it is a
ring

SD: In let “R” be a ring, I don’t think you would tag it as a ring.

NS: what about vectors? Would bold x be tagged as a vector?

SD: Good case. You might it to act like a variable (not tagged). So maybe
tagged as “bold faced x”

DF:  I disagree. In the latex source, I would write \vec{x}. Different
authors might use bold face or might put an arrow across the top.  The
semantics are the same, even though the presentation is different.

DF: I want authors to write \vec, so I want that info kept in the mathml

SD: good point. I’ll change my mind in my answer.

SD: Lines 1223 and following should be changed because there are multiple
displays for vectors.  If you want to round-trip, then the vectorness needs
to be specified.

NS: The role would allow you to recover it as \vec if it were marked as such

DF: My point was that \vec is the real thing, not that it is boldface.

SD: My framework was to allow both uses.

NS: Do you allow multiple IDs so both are needed?

DF: Still not clear why you need to say it is a boldface v.

DC: I would be much happier if every line in the spreadsheet where I know
the meaning were deleted. I don’t think they add any semantics.

SD: would it help if I said “don’t use it”

DC: No, I have to document it. They add a lot of weight for zero content

DC: I can see keyboard users wanting a name. But that’s not semantics.

SD: I found it useful to have names in the content mathml tree in my
editor. I could use those forms in the tree.

DC: I agree that you should be able to add any unicode symbol. It’s not the
same as adding semantic.

NS: Agree with DC -- I would need to support removing all those roles that
don’t do anything.

SD: I want to have a role that unambiguously disambiguates everything

DF: what about splitting the table into two parts

SD: some might want to use asin or arcsin for a name

SD: we have “subset or equal” and “propersubset”.

SD: what do we with openmath which is namespace

SD: how do we choose the names for consistency? Do we write up cross tables
from various standards?

DF: subset symbol: how do we distinguish between “subset or equal” vs
“proper subset”.

SD: yes, I didn’t get to that

DF: it is not just names, but that some names imply meanings. So be careful
about them.

NS: we should figure out what to do next

SD: I want to make sure we have all the content MathML symbols. Same for
nemeth braille

NS: I think there is agreement the non-semantic parts should at least be
moved to a separate table.

DF: You should be careful about the names. For example “<<” can be much
less than or asymptotic growth. Be careful with that name.

DF: Don’t use “timessign” when it could be multiplication or cross product

SD: In my editor, I had a “class” attribute to distinguish meaning and
separate it from the id.

DF: between 1388 and 1390, they all mean multiplication.

DF: don’t use the symbol name as the role.

SD: a subrole could resolve

DC?: that adds more complication

CP: I don’t think there is any way ARIA will allow 1,000 new names.

SD: maybe we need to use mathrole instead.

NS: we will have a meeting next week to continue this. David C will chair
if I can’t join the call.