Minutes: MathML TPAC meeting, 14 Sept, 2023

 Attendees:

   - Neil Soiffer
   - Louis Maher
   - Murray Sargent
   - Bert Bos
   - Michael Wilson
   - Sam Dooley
   - Bruce Miller
   - Brian Kardell
   - Deyan Ginev
   - Cary Supalo

<https://sandbox.cryptpad.info/code/inner.html?ver=5.4.1#cp-md-0-regrets>
Regrets

David Carlisle
<https://sandbox.cryptpad.info/code/inner.html?ver=5.4.1#cp-md-0-agenda>
Agenda
<https://sandbox.cryptpad.info/code/inner.html?ver=5.4.1#cp-md-0-1-questions-comments>1.
Questions/Comments

We introduced ourselves.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.4.1#cp-md-0-2-the-future-of-content-mathml-mathml-5->2.
The future of content MathML (MathML 5)

DG: We need to get support among other communities before we go much
further with this. Representatives of the Lean community might be
interested in this.

DG: We have a lack of consumer standards.

NS: How many people are we talking about?

DG: We are talking about around one thousand people. Not many developers
But Lean has good luck attracting mathematicians. Lean is used to prove
theorems. The algebra systems have large databases.

NS: Mathematica and Maple may not use MathML content.

We discussed the fact that standards allow many organizations to
communicate, but might sacrifice some of the organization's advantages by
everyone having access to a standard. This can happen when a new method is
discovered.

DG says we need to do more community building to see what users need from
MathML content.

NS: asked SD where SD wants to go with MathML 5 because he has a
content-MathML based editor.

SD found that MathML content works better than presentation MathML for SD's
Braille equation editor.

SD: How can we communicate the value of content MathML?

BM: We need to know if developers know about content MathML.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.4.1#cp-md-0-3-implementation-plans-for-mathml-4>3.
Implementation plans for MathML 4

NS: Who is going to implement MathML 4 with intent.

MuS demonstrated his application.

*ACTION* MuS will want to describe his application.

The application allows people to enter math by typing or by speech and
having Unicode characters being entered into a document. The application
can also generate SVG math.

People can speak math faster than they can type it.

He started developing the application near the end of the nineteen
seventies.

The application was useful at Microsoft.

The application is in the public domain.

NS: How does this relate to intent?

MuS: His idea is that his application will become more effective once we
get the list of intent names.

DG: Some of the claims for this application are too strong.

DG: We should finish the intent list and then make this application more
accurate.

NS: Intent is meant to get the speech right. If you speak it on input, it
seems like a great starting point for specifying what the intent is of the
MathML that is created.

BM: Because content is hard to get right, intent is a nice prod for us.

DG: There was an accessibility forum in June. There are hundreds of people
interested in MathML and with arXiv. This is a resource for testers.

DG Spoke about the need to have workflows for generating and consuming math.

DG: People need a workflow that will let them hear how the math they are
generating is spoken. LaTeX users need to be able to generate html and hear
how the resulting HTML is spoken.

NS then demonstrated some of MathCAT's capabilities. MathCAT stands for
Math Capable Assistive Technology.

Neil's cat made a brief appearance. He explained his cat's meowing was the
impetus for his company's name: "Talking Cat Software".

The MathCAT demo can receive math in various forms: ASCIIMath, LaTeX, and
of course, MathML. It generates both speech and Braille output. He showed
off speech highlighting and also highlighting while navigating the
expression.

MathCAT makes use intent if it is present. If intent is missing, then
MathCAT tries to guess the intent for some notations.

MathCAT is used in the production version of NVDA, and the 2024 beta
version of JAWS. It is also used in American Printing House for the Blind's
Braille Blaster and some other software.

A potential project is to integrate MathCAT or SRE (if it implements
intent) to editors LateX previewers so authors can not only see their
writing, but hear how it sounds also. For someone who is blind, that's an
important feature.

MuS can use his application to generate math from his speech output, and
that math can be turned back into speech with MathCAT.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.4.1#cp-md-0-4-what-should-we-do-about-tests-for-mathml-4->4.
What should we do about tests for MathML 4?

NS: What are we going to do about testing. This is critical for MathML 4.
How are we going to test MathML 4 which is outside the web platform?

NS: We may only have tests that can be judged by humans; that is, we cannot
write machine-controlled tests.

NS: How do we efficiently run such tests, and where should the tests be
stored?

BM: We can easily write tests with html and JavaScript.

NS to BK, if you were asked to write tests for MathML 4, what would you do?
What do you use for input, and what is your output?

MUS: has two tests: one for dictation, comparing the input an the output,
and a separate test going from Unicode to MathML.

NS: Speech output could be part of the test. Where do we put the tests
because they are not web platform tests.

DG: Start with LaTeX, generate intent, and examine the correctness of the
intent.

NS: Would like to have LaTeX input and speech output.

NS: half of the effort in developing Mathcat was cleaning up the
mess/different forms of MathML that software generates. For example, the
number 1,234 often has the MathML <mn>1</mn><mo>,</mo><mn>234</mn>

DG started a discussion on how the comma and period are used differently in
math in different languages.

NS: In Vietnam they use both commas and periods as numerical separators.
The use of commas as separators in numbers like 1,000 comes from American
influence. So it can be quite messy!

DG: What should the group be testing?

SD: Have we thought about an intent normal form? Can we produce the same
intents for different inputs?

We may have to specify tests as a function of location and language.

NS: As we learned last week, the colon and divided by symbols are treated
differently in different countries.

DG: What are our test requirements? Is it sufficient to say that we have
human-run tests for MathML 4?

NS: There are groups dealing with software that are not web based. What do
they do for testing?

NS: Philippe Le Hégaret said we can have human-run tests; but we must have
tests.

NS: It is better to have machine-run tests than human run tests.

BK: gave an example of testing but was not audible.

*ACTION* NS said that we should meet with the ARIA group to discuss
testing. They meet at the same time as the Math WG group meets.

BK praised the demos.

SD: is interested in generating presentation MathML with intent from
content MathML.

SD wants to use Braille as both input and output.

NS: how does SD's work fit in with everyone else's work?

SD said he had several synergies with MathML 4 because he wishes to
transform Braille input to presentation MathML with intent and back to
Braille. He will use content MathML in this process.