# Minutes: MathML Semantics meeting July 16, 2020

• From: Neil Soiffer <soiffer@alum.mit.edu>
• Date: Thu, 16 Jul 2020 16:40:41 -0700
• Message-ID: <CAESRWkCBWkCdepuYH3tY-9Sfq_jf8K76CCB0AYSM7Buw8RAEUA@mail.gmail.com>
*Several people contributed to the minutes, but there are several gaps.
You'll need to listen to the recording to get all the details. The meeting
was recorded:
https://benetech.zoom.us/rec/share/uPRsLuv17jtOYZX1s3mEfZ4NGb31eaa82iIWrKIEn00fT-Sh1DKryoVfqCshBuo4
<https://benetech.zoom.us/rec/share/uPRsLuv17jtOYZX1s3mEfZ4NGb31eaa82iIWrKIEn00fT-Sh1DKryoVfqCshBuo4>
password 4O$A$26cAttendees:Neil SoifferDavid FarmerDeyan GinevSam
DooleyLouis MaherDavid CarlisleMurray SargentPatrick IonBruce
MillerContinue discussion on semantic proposalsNeil to report on his
attempts to use "pure" semantics for speechNS: started with Sam’s notation
-- semantic=”binomial”, arg=”1”, arg=”2”NS: easy to deal with -- when
writing a rule to consume it, I know the names. Here’s a sample rule:any ?
(semantic=="binomial")=> UIWord(@"1",@"2"){ruleRef="RR_binomial";
simple=IsNode($1[0], "Simple") && IsNode($1[1], "Simple");};NS: basically
is says for “any” tag that has the attribute semantic=”binomial”, generate
a template that grabs the element with attr=”1” and “attr=”2”. No need to
figure out their names.NS: if were semantic=”binomial(@upper, @lower)”, the
code would still need to refer to the first and second arguments, but now
the internal matching code would need to parse the value of ‘semantic’ and
then get the first arg, see what it is called, and then get it.NS: problems
with Sam notation for \int \frac{dr}{r} SD: could coin a new name
‘intInFrac’DF: Stieltjes integral is d(function).
https://en.wikipedia.org/wiki/Riemann%E2%80%93Stieltjes_integral
<https://en.wikipedia.org/wiki/Riemann%E2%80%93Stieltjes_integral> Weighted
integrals (orthogonal polynomials) are another example, where you have
weights like \frac{dr}{\sqrt{1-r^2} . So a good alternative name for those
types of integrals is important anyway.[PI: In the example case, we
actually have d(log r) = dr / r .]NS: for nary, would like to have the
semantic on the operator so I know how to pronounce it more easily.BM: I’m
all for simplifying notations as much as possible, but no simpler. What is
missing is the thing itself and the thing applied to arguments.BM: gamma^*
is some incomplete function, but can be applied.SD: So an
semantic=”gamma-star” would be on an mrow or on an msup.BM: It should be on
the msupNS: Wouldn’t you have semantic=”function” on the mrow and
semantic=”gamma-star” on msup.SD: An alternative is to markup semantic on
the mrow.BM: need to distinguish between the function and the function
applied to the arguments.[free ranging discussion about operators,
functions, naming, etc -- listen to the recording starting about 25]*

*NS: work on this doc
https://mathml-refresh.github.io/mathml/docs/comparison
<https://mathml-refresh.github.io/mathml/docs/comparison>Progress/issues by
others on generating semantics or consuming itDG: I did more work
validating out work. Overall it works quite well with the DLMF where math
is mostly K-12.DG shows an example of of x_i^’ written in different ways.
Derivative with implicit variable. Other examples…SD: msupsub is definitely
something that needs to be detangledDF: also \bar{x}_i even though the
semantics are x_i with a bar over it.PI: Translation is a different
matter;  I’ve translated books from French and German for Springer and
you’ll need enough recorded information in the markup to do a good job.
Translating from old MS word is already much more difficult than from
LaTeX.Comments on different types of integrals (from David F)Not all
integrals are of the form\int f(x) dx .There are (at least) two other
classes:1. Stieltjes
integralshttps://en.wikipedia.org/wiki/Riemann%E2%80%93Stieltjes_integral
<https://en.wikipedia.org/wiki/Riemann%E2%80%93Stieltjes_integral>These
look like\int f(x) d g(x)If f and g are sufficiently nice then that
equals\int f(g) g'(x) dx ,but the Stieltjes integral is defined in more
general cases.2.  Weighted integralsThe Chebyshev polynomials T_n are
orthogonal with respectto the inner product\langle f, g \rangle =
\int_{-1}^1 f(x) g(x) \frac{dx}{\sqrt{1-x^2}} .It is common to write that
weight as a fraction:\frac{dx}{\sqrt{1-x^2}}Some times the weight looks
like part of the integral,but it actually serves a different purpose which
may be helpful toindicate.  For example, the Hermite polynomials
areorthogonal with respect to the inner product\langle f, g \rangle =
\int_{-\infty}^\infty f(x) g(x) e^{-x^2} dx .Based on the above examples,
it seems we need a way to indicate themeasure in the integral (for those
cases where the measure is notjust "d of the variable"), and a way to
indicate that part of theintegral is a weight.Moving towards a working
groupNS: We need to move towards a working group so that the Core work has
normative references. The W3C has virtual meetings (TPAC) in mid October
followed by “breakout” sessions at the end of October. We should have a
charter ready in early October to get early feedback from various groups
before moving to the formal application. The charter among other things
lays out work items. We need to start the process of figuring out our
deliverables, some time lines, and work requirements for membership in the
group. I will schedule a general meeting in the next two weeks to gather
bullet items.DC: Next Weekend is TUG2020 some talks on accessibility:
https://tug.org/tug2020/ <https://tug.org/tug2020/>*