Minutes: MathML Semantics meeting Aug 20, 2020

Very lively conversation today and the time went by quite quickly.

The meeting was recorded:
https://benetech.zoom.us/rec/share/9-5bN77LyX5JZ53CskrxSIgNMMPHT6a80XMe-fUOy033OoNzs_1V2VeVeyiJQtnB
Password: eB2L+%pb (starts about 5 minutes in)


Attendees:

Neil Soiffer

Deyan Ginev

Louis Maher

Murray Sargent

Bruce Miller

David Farmer

David Carlise

Steve Noble

Patrick Ion

Regrets: Moritz Schubotz


Charter is at
https://docs.google.com/document/d/1W-oYUbOMueaqb3KFSWkjWVBwR6AzSEBizHwQhvSwfDc/edit#

1. Last call -- charter comments, suggestions, discussion. Especially goals.

Charter is at
https://docs.google.com/document/d/1W-oYUbOMueaqb3KFSWkjWVBwR6AzSEBizHwQhvSwfDc/edit#

NS: Alignment seems to be a problem in MathML 3. There are potentially
additions that could be made to MathML 4. Currently many people use tables
which messes with semantics. Should the charter call this out?

PI: I think alignment is a basic part of layout

DC: Seems like too detailed. Might be simply part of semantics.

MS: I’d like to know what we did wrong in our implementation of
malignmark/maligngroup.

DC: There are a lot of complications

MS: We probably don’t support all of them.

NS: Maybe we need to simplify them if they are hard to implement

SN: Pearson is very concerned about elementary math accessibility and
braille

NS: There is a polyfill, but probably only MathPlayer speaks it. Braille is
an issue because it is 2D

LM: The limited width (40 cells) is a problem

LM: Same for systems of equations. Working like a table would be best so
one can traverse columns.

NS: That’s what MathPlayer does, but doesn’t handle malignmark/column

MS: Office doesn’t do it yet.

BM: We should think about allowing semantics outside of MathML, for example
on a table

NS: I think it is a very heavy lift to get other groups to accept that.

BM: I agree with DC that we should not commit to it in the charter but
saying we will “examine” it is just weaseling.

PI: I agree also since semantics might just cover it. Could try parallel
markup, but it would be complex.

MS: Search needs to ignore the splits.

General agreement: let’s not add that to the charter specifically.
2. Semantics names -- continue discussion   a) Anyone notice any general
groups missing?

DG: We should add an alias column so people searching for a name will be
able to find whichever name they prefer using.

DG: Maybe there should be a separate sheet for constants vs “types” / other
things. An example would be large tuples that represent some mathematical
structure. Say an algebra, logic, group ring with a carrier set (or
several) and some operation(s). They are kind of out of scope but still
useful for e.g. search and e-learning applications (where one can “ask”
what the tuple meant). I’m not proposing we use them for the core a11y
levels, but they do keep cropping up. We need to make some decisions on
them / discuss …

NS: let’s have a meeting focused on this topic with
   b) What should be removed (some are clearly not appropriate)?

(skipped)

PI: What is ‘z-omit’ in subject column in level 1?

NS: They are things that Sam and I found from content MathML that probably
don’t go into semantics value. Things like ci, cn, …

NS: Some things like bvar are more questionable because Sam used them in
presentation MathML he generated.
   c) Names -- naming scheme and do we want to keep some content MathML
names (e.g., "lt" or spell out as "lessthan" or "less-than")?

NS: the content MathML names are not consistent. Some abbreviated (‘lt’ and
‘card’) some not (‘cartesianproduct’).

DG: I like long names with ‘-’ so they can be pronounced easily. E.g.,
“solar-mass” and “earth-mass”. Makes lookup easier in wikipedia and other
places. It also encourages practitioners (e.g. high school teachers) to
consider contributing names that may be useful for their materials.

NS: I did experiments with speech engines. Neither CamelCase, hyphens, nor
underscores works well to be pronounced across all speech engines. Probably
need to change to spaces if they are to be spoken.

NS: Some of the rationale for naming to “spell it out” doesn’t make sense
for common things like “gcd” (not “greatest-common-divisor”); “sin” not
“sine”. Are you advocating for a purest point of view of always spelling it
out or are you suggesting there are exceptions?

DG: I’m advocating for a primary form to be a soft standard. The aliases
resolve to it.

NS: I thought you wanted aliases to be a means to lookup up the proper name
to use. Are you saying that an alias should be recognized by all the
software?

DG yes, like wikipedia, and all names are “technically” viable (e.g.
basically redirects to a main name)

NS: I think that is terrible because there is a ton of equivalents so all
consumer’s job is more complicated. Eg., a search on “lt’ needs to match
with all possible other aliases.

BM: I like DG’s idea.

DC: I don’t see the point of naming every possible “mi A”.  If we are
talking about speaking the names, I don’t understand why we have all these
level 3 names. Just let people add whatever they want it to pronounce as.

DG: I think the importance is that level 2 and 3 are invitations to add
more names and not just the level 1 names everyone supports.

DC: but people just name things the way they want: the Ackerman function,
Ackerman’s function, the function of Ackerman.

DG: [post meeting] “ackermann-function” would gain us 26 translations for
internationalization by e.g. grabbing the wikidata reference names:
https://www.wikidata.org/wiki/Q341835

DG: The various language declinations are out of scope, and choosing a
“base” variation is possible (in fact mostly unique). “Ackermann function”
is a base, possessives (‘s), determiners (a/an/the), plurals (functions),
etc are out of scope for my proposal for the list.

DG: Level 2 and 3 are invitations to add more names. Addresses my gripe
with OpenMath. People are afraid to add more.

NS: one would be an alias?

DG: no, aliases are reserved for alternative names, abbreviations etc.

[discussion of pronunciation of math

DG: There is very little confusion in terms of language. Possessive form
doesn’t belong, “the …” doesn't belong. Just want dashed name. Way to think
of table is to think of it as definitions and the “known notation” field
connects that definition to a presentation. Choose one base.

PI: Euler’s constant vs Mascheroni’s constant?

DG: yes, precisely

BM: which would we use

DG: Either, both “technically viable”, and likely one would be primary
based on community agreement.

DG: An additional problem with my suggestion, however, is that aliases will
lead to collisions when you use natural language (same name used for two
constants/concepts). Consider wikipedia’s “disambiguation pages” which are
born from the same problem.

3) Continued discussion on "semantics"

[didn’t get to this]
   a) discussion of David C's idea to deal with alignment
   b) continue discussion on n-ary operators, especially +/- and relational
operators