Minutes: MathML Full WG 17, Oct 2022

   - Neil Soiffer
   - Louis Maher
   - David Carlisle
   - Bruce Miller
   - Neil Soiffer
   - Deyan Ginev
   - Sam Dooley
   - Moritz Schubotz
   - Paul Libbrecht
   - Steve Noble
   - David Farmer
   - Cary Supalo
   - Murray Sargent

<https://sandbox.cryptpad.info/code/inner.html?ver=5.2.0-rc1#cp-md-0-regrets>
Regrets

   - Dennis Mueller

<https://sandbox.cryptpad.info/code/inner.html?ver=5.2.0-rc1#cp-md-0-agenda>
Agenda
<https://sandbox.cryptpad.info/code/inner.html?ver=5.2.0-rc1#cp-md-0-1-announcements-updates-progress-reports>1.
Announcements/Updates/Progress reports

NS: The mhchem package does not use Unicode chars for three types of arrows
indicating types of equilibrium. Unicode does not have the char points so
Murray and I will submit a proposal to add them.

DC: will forward an email address of another TeX chem users to NS so that
NS can report any other missing chem symbols.
<https://sandbox.cryptpad.info/code/inner.html?ver=5.2.0-rc1#cp-md-0-2-paul-39-s-a-href-https-w3c-github-io-mathml-docs-spoken-math-survey-spoken-text-survey-questions-a->2.
Paul's spoken text survey questions
<https://w3c.github.io/mathml-docs/spoken-math-survey>.

PL: summarized his survey proposal. People wanted to know how math
expressions are spoken in many languages. We want to be able to generate
intent values that would allow math expressions to be spoken correctly in
many languages.

One of the TPAC committees suggested that we generate about 100 math
expressions and see how they are spoken in many languages. They suggested
that we use a survey for this purpose.

We must find people who are bilingual in various languages so that they can
communicate with our WG. We especially need people who are bilingual in
Arabic and in Asian languages.

NS: When we get a final list of math expressions, we will write the intent
that would allow the expressions to be spoken in English.

NS: We should concentrate on expressions that are ambiguous. Intent is used
to make ambiguous expressions verbalize correctly.

BM: Once you get anyone involved in the survey, you want to get as much out
of them as possible. For this reason, we should have our expressions worked
out before beginning the survey. We should have completely annotated MathML.

MOS: We should try to minimize our intent elements. Things clear in one
language may not be clear in another language.

DG: We need to define our defaults before we start this survey so that we
clearly state what we want.

NS: Unicode characters should be supported by the math to speech system in
all languages. Some letters mean different things when their font changes.
An example of this is the letter 'I' which might have a different font when
it is indicating an imaginary number.

DG: What about abbreviations like sin and cos? How do different languages
deal with them?

NS: hopes that math-to-speech engines would already handle abbreviations.

NS: We do not know if Asian languages use English characters for
mathematics or other characters specific to those languages.

MUS: uses 18 languages for his engine. He is worried about word order
changes.

PL: The letter 'a' might be pronounced differently if it were considered to
be a set or an element of a set. The same character might be spoken
differently depending on how it was used.

DG: If we have not specified our defaults, we are not ready to consider
other languages.

NS: Symbols, used to indicate units, should be spoken clearly. Unless you
have intent, or clear context, you do not know that the letter 'l' should
be spoken as liters. We really need to specify what we expect.

BM: The survey must focus on complete implementations.

MOS: Can MathCAT generate intent values so that we have something to work
with? Can we use MathCAT to see where intent is needed?

NS: If an expression does not have intent, then MathCAT could use
heuristics to speak it. Another approach is that if an expression does not
have intent, then MathCAT could speak it just as it is written.

MOS: If no intent is given, then use heuristics to speak it. We could
improve the heuristic values by comparing how a human would write an intent
for the expression.

DG: Use NS's process to get out intent. The final step is to output the
narrated stream. You will know if you need to fix something when you
examine the full narrative stream.

DG: The authors want to see intent so that the authors know what to do.

NS: We have absolute value. Put intent on it. This is simple. Often what
gets generated is not a simple mrow structure, other variables are sucked
into the absolute value verbal description.

If we have the absolute value of x + y, it is not clear how this expression
would be spoken in other languages.

How can we mark an expression so that you know when the operation ends.

DG: Can you use the length of the expression to know when to speak the end
of the expression? For example, the absolute value of x + y should say the
absolute value of x + y close absolute value.

DG: Can you use the length of the argument to know where to close it?

BM: We discussed this last week when we talked about tables. It depends on
functional notation. If you have expressions laid out with mrows, then your
intent can be simple in most languages. It is when you do not have enough
mrows to encapsulate the expression, you have problems.

NS: The problem is that there are not enough mrows. You would have to use
intent to cover this lack.

PL: We could request that the survey responders tell us how intent would be
specified in their languages.

DF: What do the survey participants do when they encounter an ambiguous
expression?

PL: The survey people should give their best opinions in their own
languages.

NS: How can people know what vertical bars mean? It could mean cardinality
or absolute values.

PL: They must use context if possible. We must give context, so people know
how to speak something.

DG: Show someone "A superscript t" and that person would say transpose.

NS: We must give context if we expect an answer.

NS: Some languages do not say left and right but east and west. This may
not be the case in math.

DG: We should stick to the intent core.

PL: wants to use ambiguous expressions.

NS: The survey is important so that we can satisfy the needs of
internationalization.

NS: There is one more week before we start changing time zones. We will
keep the American time zones for our schedule.

DG: sent out an email listing several MathML-to-LaTeX translators. Do
people know of other translators?

MUS: MS word can do translations.

NS: Encourages email to continue the discussions.