Re: Reminder: MathML Intent meeting 2 Feb, 2021

Hi Neil, all,

That's a fair summary. For those interested in details, here's a list of
related observations that working with our group seems to have clarified
for me:

1. The "intent" level 1/2/3 lists have a lot of similarity to the wikidata
demo Moritz gave, and as resources can be seen to "induce" CMML content
dictionaries (if one wanted that).

2. We can't assume CMML trees as a given for an arbitrary expression
covered by the spec. As David Carlisle had elucidated, we have no guarantee
of any content tree at all when starting with arbitrary PMML. Additionally,
partial annotation is a design goal to make life easy for
authors/remediators.

3. As far as serializing a content tree in parallel attributes, Bruce's
"semantics-mini" proposal was already a good demonstration that we could do
that if we wanted to.

4. I still think there is something to win with Sam's approach to establish
a "Defaults" add-on, which could auto-remediate raw PMML that is commonly
used in standard K12 math. That add-on *may* also induce full CMML trees at
level 1, still to be determined.

5. We've mentioned on several occasions that certain content
interpretations rely on near-canonical PMML ("correct" mrow structure,
"correct" invisible operators, etc), and I see that as tied to discussing
mathematical "notations". A Defaults proposal that wants to standardize
auto-remediating CMML trees will inevitably need a well-defined grammar of
notations, to go along with the level 1 names.

---

My inclination would be to check the "Defaults" approach in practice - grab
a chapter from a real book and see how many of its equations we can fully
match against a "K12-defaults" set of notations, in the process refining
both the level 1 names and our tree-building rules. If the notations are
clear, we can induce both better speech (where needed) and a full content
tree, but whether there is a "reasonable enough" set of defaults is a math
linguistics question more than anything. E.g. for arXiv there is no
reasonable closed notation set. For a textbook on highschool trigonometry,
there might be.

Greetings,
Deyan

On Wed, Feb 3, 2021 at 1:34 AM Neil Soiffer <soiffer@alum.mit.edu> wrote:

> We meet on Thursday, 10am Pacific, 1pm Eastern, 7pm Central European Time.
>
> We will be using the zoom link from last week. If you would like to join
> and don't have the link, please send me email at least 10 minutes before
> the meeting.
>
>    1. Announcements/updates
>    2. What's missing from Level 1?
>    3. Start discussion: what's the intent of 'intent'...
>
> We've spent the last couple of meetings discussing level 3 ideas. I think
> we have neglected level 1. In looking through some speech rules, I see a
> number of items that aren't covered by level 1 (or level 2) that I think
> belong. Please take a look at the level 1 functionality
> <https://docs.google.com/spreadsheets/d/1EsWou1K5nxBdLPvQapdoA9h-s8lg_qjn8fJH64g9izQ/edit#gid=0>
> and see if you see anything missing that should be discussed.
>
> The last item on the agenda is something that I'm sure will provide some
> lively/thought provoking debate for several calls: what do we want the
> 'intent' attr to convey? We've tacitly(?) assumed it allowed software to
> understand what the author intended at the level of being able to be
> converted to something like Content MathML (CMML). But Deyan and some
> others have suggested that accessibility should be the primary concern. We
> have seen a few examples where the current intent proposal is structured
> for CMML and not accessibility (integral example with variable of
> integration in the numerator). What other notations are like this and does
> it matter?  intent="integral(...)" doesn't really tell AT anything that
> isn't already obvious since there is only one use for an integral sign.
> Maybe that's not the case with some other examples. Perhaps there should be
> a prize for the most divergent case of speech and CMML? :-)
>
>
>
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> Virus-free.
> www.avg.com
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> <#m_-5712169669025599581_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>