Minutes_ MathML intent meeting 27 Jan_ 2022

Attendees:
* David Carlisle
* Sam Dooley
* David Farmer
* Deyan Ginev
* Louis Maher
* Bruce Miller
* Moritz Schubotz
* Murray Sargent
* Neil Soiffer
* Steve Noble
* Bert Bos
* Paul Libbrecht
* Patrick Ion
* Cary Supalo
* Stephen Watt

Regrets:


Announcements/updates

There will be a MathML Core meeting on Monday.

ClearSpeak's "preferences" to resolve intent vs speech? [I might get this written up as an issue before the meeting -- I will send email if I do]

NS: The Educational Testing Service (ETS) ran tests to see how blind people would like mathematics spoken.

NS: There are preferences in MathType so that the listener can have some control over how math is spoken.

DG: Was there a conclusion to see how people like to hear things?

NS: No conclusion was reached on how math should be spoken in their ClearSpeak tests.

NS: The ClearSpeak preferences are well documented.

NS: ClearSpeak:
https://docs.wiris.com/en/mathtype/office_tools/accessibility/clearspeak

SN: MathJax has speech preferences.

PL: The speech tests are designed by psychologists who may not have the same views as mathematicians on how math should be spoken.

SN: There were user studies using blind individuals on how math should be spoken; however, there were not enough subjects to get statistically meaningful
numbers.

NS: ETS said the clearSpeak voice output was preferred by students over just reading what is there.

SN: The gold standard would be a large international study with many subjects. This has not been done.

DG: How much priority should you give to clearSpeak output instead of making the intent work to disambiguate symbols. Making the intent work is more important
than controlling the speech style.

DG: Can we use the CSS dialect?

Progress on aliases for level 1 names? --
https://github.com/w3c/mathml/issues/257

LM: Alias names can be called "also known as" names.

SN: The terms "related" and "supplemental names"" were also suggested to replace "alias".

DF: Discussed the terms "is an element of" and "in", used in set definitions, as examples of giving different names for the same symbol. A person experienced
in math might want to hear "in" instead of "as an element of" when discussing a set.

DF: Who should control the speech, the listener, or the author? CSS lets the author win.

DF: Did not like "also known as". Sometimes terms have extra meanings that keep them from being perfect aliases.

MUS: The user should have a major role in verbosity.

NS: has added relative speech speed to MathCAT.

LM: listens to math at the normal reading speed of around 150 words per minute. He does not speed up the reading speed for mathematics.

SW: Most of the time, authors do not concern themselves on how their math will be spoken. Many symbols contain intrinsic ways to be spoken.

NS: Authors do know how they want things to be read.

SW: Authors are mostly uncaring about how things are spoken.

BM: Because words can have extra meanings, authors care about how things should be spoken.

BM: We should decide which questions we are trying to answer. Maybe we should pay attention to the author, and maybe not.

NS: Is it useful to have multiple names so that an author, or listener, can have a say in how things are spoken?

PL: He is not clear on the purpose of deciding what term should replace the word "alias".

PL: While an author might not care about speech, the publisher wants to have high quality speech.

NS: With Intent, if you want to say something other than factorial, use a different name.

LM: Are you trying to guide speech by entering words or by selecting from a table of words?

DG: Cardinality, length, and size are the same thing.

DG: Cardinality would not be introduced in middle school, but size would. You want people to know that they have the same meaning underneath. Let the author
write what he has in mind.

DG: The main role of intent is to disambiguate the name. Let an author write the name they know.
Level one names are spoken as they are written. There is a set speech definition for level one names.

SW: People can make spelling mistakes which make speech systems fragile.

BM: Level three is always there for people who want to introduce new things.

BM: The author does not need a high level of speech control.

NS: There are different ways to say the same thing. For example, "one over three" or "one third".

dg: For power we could say exponent.

PL: We have not converged on naming. We need to do something practical. Write down several symbols and see what speech would be clearer. We do not know
what to do about aliasing.

PL: In French and German you have several ways to speak about fractions and powers.

PL: If internationalization should enter this discussion, the problem would be much more complicated since individual languages have their own aliases.

NS: We will have to have a set of aliases for each language.

DG: No matter what language you use, the aliases for power would map back to the concept of power and the computation would be done correctly.

NS: Automatic translations are not dependable for this process.

DG: The table is supreme. All users must look up the correct term in a table; therefore, the table rules.

DF: The same string of symbols could mean absolute value, or determinant, or cardinality. I see our main goal as providing a way to disambiguate which it is. Once we know what it means, there may be more than one way to say it. It is asking for trouble if we use intent both to disambiguate and to specify
a preference for pronunciation. Maybe we could use a different attribute to indicate pronunciation, if we decide it is worth trying that.

BM: If we are only talking about the core, where all is known, the author does not need to control speech.

SW: It is not the case that we always know what symbols mean.

DG: The absolute value is also known as modulus.

BM: The table controls the speech.

DG: The author should control the speech.


Regards
Louis Maher
Phone: 713-444-7838
E-mail: ljmaher03@outlook.com