Minutes: General/semantics meeting on Oct 15, 2019

Meeting was recorded:
https://benetech.zoom.us/recording/play/l484nMxeil-cOIVM94sK7IRXofjazeJ_iW9_h5Bf2-w2A_JNzj56vspf-We6dIyF?continueMode=true

Attendees:
Neil Soiffer
Steve Noble
David Farmer
Charles LaPierre
Sam Dooley
David Carlise

Patrick Ion

Discuss #141

SD: review of what was covered last time

Need something like aria role but for math specifically that can be put on
a MathML element that applies to the entire tree for that element. To add
semantic markup to identify on any presentation mathML element what the
meaning of that element is.

SD: two use cases: mark symbols etc so one can speak what the symbol means

SD: other cases: convert to content MathML, so mark everything, not just
“weird” things

NS:  Do you need to map new items to some predefined name?

SD:  If there is something you don’t recognize, it is just an undefined
symbol.

SD: this is the doc I wrote:
https://docs.google.com/spreadsheets/d/1ebOkl7Gckfk5g6Dc4C8bpGZtSxLnGwpOHqAwwON0-nI/edit#gid=135098649


DC: if you’re annotating presentation you know what the character chosen
is. The meaning, say as times, is separate.

DF: Integration by parts is being discussed,say, with a  dot to denote
multiplication to emphasize.

Do you want to be able to emphasize that?

SD: ...

NS: you know it’s a times dot.

DF: the ‘meaning’ is just times, not times-dot.  I don’t want to put both
meaning and appearance in the 'role'

NS: there are 4 times in there; times-dot, invisible-times, …

I’m not sure about plus-minus.  I’d suggest the role is the Unicode value
for plus-minus

SD: all examples in column E are for illustration only

E.g. ‘negative’ used for negative numerals with superscript minus sign. ….

NS: Aria people didn’t want hundreds of roles

SD: only use roles if you want to do math editing; the more we name the
better; but there has to be a certain amount only; line 490 ff has a group
of things we could drop

NS: if the semantics are inherently described by pmathml you don’t need
roles in addition then

SD: that builds in a tie between semantics and presentation; I think of
generating programmatically

NS: issue is size of that list


NS: one of goals of MML 4 should be software interprets the same MathML in
the same way; and mfrac without a math-role should have a meaning agreed
upon (and specified).

DC: you’re embedding content in pres;

SD: Yes, I am

DC: shadow DOM is to be implemented anyway

NS: cf link to Fred’s talk just given with an example of use of SDOM and
“la-tex”; I suspect that could be done with pragmatic CMML.

SD: that would be incredible to have an implementation of CMML using SDOM.

NS: DF in PreText you have a number of semantic macros for mult but not for
plus or

equals; how did you draw a line?

DF: I listed at the start those things I thought already semantic enough;
and that includes,

+, - and exponentiation; nobody is going to write this directly



DF: my doc is
https://docs.google.com/document/d/1cZnff5_fi_ucNyZ1ex2msmJLE55FAZD-QInkLYe8xiEly


DF: I want to create MML with extra semantic info when it’s available;
Pretext is just an example; but the sample book is used by a lot of
students at many places

I change a small change in your habits; at end there are lots of examples
of small changes that you can do an express the semantics better; I wrote a
Python script to process the input into a more complex macro expression
(and that outputs extra markup that authors don’t see immediately) .
 <longer careful explanation by DF>

DF:  The python script is just for proof-of-concept.  I am not sure how
this would be done in practice.

[PI: try Controlled Natural Language after Tom Hales:

https://www.icms.org.uk/downloads/bigproof/Hales.pdf]


DF: what can I do now to help?

PI: what you have done is a very good start.

DC: agree, but I don’t think we necessarily want to use those names.

DC: very similar to existing current TeX->MathML converters. But those lose
the semantics. Useful for PDFs also. Lots of work in making that PDF
semantic.

DC: it is great for informing us what roles to use

PI: if you have TeX, you use TeX

DF: I did this as a proof of concept. I wanted to see how much work was
needed for calculus to add semantics. May do a discrete math textbook next.

PI: I agree with DC that we aren’t trying to standardize the TeX input.

PI: Do we really need to list the accepted strings to use for roles?

SD: If we leave it open-ended, but have a list like in content MathML, that
gets us pretty close

PI: but you need to know what the accepted list is

DC: in practice, it has to be open-ended because no validator will object.

PI: I fully agree it should be open-ended. We can take a very lax view on
what is allowed.

PI: I think it might be good to allow a list of semantics. I think aria
roles are too narrow.

SD: I created my mine based on feedback from various groups at Pearson.

DF: are you going to change to times from times-dot

DF: I’ll send you my list to add to your spreadsheet

The next meeting will be in three weeks:  5 Nov.