Re: [EXTERNAL] Notable mention of rich math applications, "details-on-demand"

Hi Moritz, all,

Happy to see the article got everyone energized and thanks to all for
the related links! Very interesting to see other cases of colorized
math+text, and to see an example of the wikidata Content Dictionary
approach again.

Indeed the current MathML 3 can be used to build this
"details-on-demand" demo in theory, and has been used for similar
demos in the past, as I mentioned in my original email, including
projects such as DLMF and Planetary.

Your wikidata Content Dictionary with cross-referenced P+C MathML is
certainly one of the more "freeing" ways of using CMML, as you can
reside in a single CD with all symbols you'll ever need. Cool! Having
a way of cross-referencing the presentation and content markup is
indeed a great boost, even though it seems people continue to
discriminate against programming languages, so you end up needing
"cross-referenced presentation and content MathML generated by the
Right programming language".

As you mentioned the "legend" part of this application is natural
language, so it has to be added on top of the MathML spec. Which is
fine, and easy to do if you have ids and a hashmap that connects each
id to its natural language label. Ideally you'd connect to ids in the
Content tree, and have them colored via their xref into presentation.
Doable, and even gets n-ary operators highlighting correctly, if
xref-ed correctly - something Bruce had shown me in the past.

Content MathML is generally expressive enough to carry most
applications on its back in theory ("you just need a better framed
Content Dictionary and application X will work"). But it is a painful
and cumbersome vehicle in practice. CDs were never made easy to create
and reuse, so they remain hard to create and reuse. From one
standpoint, sure - we could be focusing the group on creating an
"accessibility" CD, and having cross-referenced presentation and
content trees on which to base a11y. I have been tempted to create
that demo as an alternative a11y showcase exactly to draw this
comparison - and maybe I should? latexml has been able to do the hard
representation work (xref-ed P+C MathML) for some years now, so it is
something I can do on a short notice...

But the sentiment on the calls has been quite clear to me - a lot of
aversion to using 1) two parallel math trees, 2) cross-references,  3)
global ids, 4) CDs and 5) generally any data structure that isn't
immediately needed by the applications we are discussing. In other
words, we have a minimalist sentiment, and want to create something
very easy to use that can solve its core problem. Accessibility seems
to be quite interesting since it uses a lot of the presentation
information for speech, but it also uses a little bit of content at
just the right places - to create compositional language with the
right phrases and connectives, and to "speak to the point" where
appropriate. It would have been the ultimate demo for cross-referenced
content and presentation MathML, but it is also the ultimate demo for
why we need something simpler. Allowing for "partial annotation" and
avoiding the second math tree are immediate wins in usability, as
people can do minimal changes to existing sources to make them
accessible. Same with avoiding CDs and letting people annotate what
they already know how to say. And we might find more easy wins yet...

Greetings,
Deyan

On Mon, Sep 14, 2020 at 2:12 AM Physikerwelt <wiki@physikerwelt.de> wrote:
>
> Hi Deyan,
>
> a nice example. However, can't we represent that with the current
> MathML standard.
>
> For Wikipedia, we have the semantic annotations, e.g., by clicking on
> the first formula in Mass Energy Equivalence you will be redirected to
>
> https://en.wikipedia.org/wiki/Mass%E2%80%93energy_equivalence
>
> or
>
> https://de.wikipedia.org/w/index.php?title=Special:MathWikibase&qid=Q35875
>
> or
>
> https://ru.wikipedia.org/w/index.php?title=Special:MathWikibase&qid=Q35875
>
> and so on depending on your language preferences. Of course, this is
> supposed to become a popup, but the code review did not go through
> since the js code was a few bytes too heavy.
>
> Eventually, we can translate this into content MathML.
>
>
>     <annotation-xml encoding="MathML-Content" id="e11">
>       <apply id="e12" xref="e2">
>         <eq id="e13" xref="e4"/>
>         <csymbol cd="wikidata" id="e14" xref="e3">Q11379</csymbol>
>         <apply id="e15" xref="e5">
>           <times id="e16" xref="e7"/>
>           <csymbol cd="wikidata" id="e17" xref="e6">Q11423</csymbol>
>           <apply id="e18" xref="e8">
>             <power id="e19" xref="e8"/>
>             <csymbol cd="wikidata" id="e20" xref="e9">Q2111</csymbol>
>             <cs id="e21" xref="e10">the integer number two</cs><!--
> this was somehow a standard violation to demo clear text in the popup
> -->
>           </apply>
>         </apply>
>       </apply>
>     </annotation-xml>
>
> However, not in production since we rely on LaTeXML, and we can not
> run PERL in production WMF sites. We, therefore, use the 'has part'
> property of the related Wikidata item
> https://www.wikidata.org/wiki/Q35875.
>
> For more complex formulae like
>
> https://en.wikipedia.org/w/index.php?title=Special:MathWikibase&qid=Q1899432
>
> , the problem is that there were missing symbols, thus we had to
> create new symbols
> https://www.wikidata.org/wiki/Q85397895 with unfortunately very few
> semantic connections to more general symbols, and no translations yet.
> Thus these annotations are almost dead ends in the knowledge graph.
>
>
> Greetings
> Moritz
>
> http://moritzschubotz.de | +49 1578 047 1397
>
> On Mon, Sep 14, 2020 at 6:13 AM Neil Soiffer <soiffer@alum.mit.edu> wrote:
> >
> > Deyan,
> >
> > I really like the toggle notation/variable feature. I think it is better than pop ups for this purpose because you can see the context better. It is something that widespread support of MathML will enable because then people can count on where something is and could do text layout around the math and have the other end of the arrow actually point to somewhere near the math symbol/term and know it will work in all the browsers. When the math input gets converted to something else such as spans, you can't do that.
> >
> > I'm not a fan of the colorized-math-equations, but maybe people who see colors better than I do would like it. I find it way too busy, distracting, and hard to read.
> >
> > To follow up on Patrick's link for Euclid's Elements. A friend of mine did something for a "live" (colored) version of Euclid's elements where you can play around with the diagrams. He published in the Apple App Store. It's not free though (it's $6). I've played with it some during development and found it quite interesting.
> >
> >     Neil
> >
> >
> > On Sun, Sep 13, 2020 at 12:28 PM Patrick Ion <pion@umich.edu> wrote:
> >>
> >> Thanks, Deyan, for bringing a modern document form to
> >> our attention, and not just for the math relevance.
> >>
> >> I can't resist bringing up a bit of history in regard to
> >> colored math.  A significant earlier work was
> >> Oliver Byrne's version of Euclid (which I saw
> >> in Cork), written while he was Surveyor of the
> >> Falkland Islands,
> >>
> >> https://www.c82.net/euclid/
> >> see also
> >> https://www.princeton.edu/~graphicarts/2008/05/euclid_in_color.html
> >>
> >> Note: it really gets going with Book II in using
> >> colored symbols.  You can buy a copy these days:
> >>
> >> https://www.amazon.com/Oliver-Byrne-First-Elements-Euclid/dp/3836544717
> >>
> >> David Joyce (https://mathcs.clarku.edu/~djoyce/)
> >> did his early Java applet-based Euclid with colors in the diagrams
> >> https://mathcs.clarku.edu/~djoyce/java/elements/elements.html
> >>
> >> Then there was the idea discussed years ago at a Paris TeX
> >> meeting of using color to indicate, say, for a Lie group $G$,
> >> its Lie algebra as $\color{green}G$, its enveloping algebra
> >> by $\color{red}G$, its representation ring by $\color{orange}G$
> >> and so on.  I may have seen this style actually used in practice,
> >> but cannot recall a definite reference at present.
> >>
> >> All the best,
> >>
> >>    Patrick
> >>
> >>
> >>
> >> On Sat, Sep 12, 2020 at 9:44 PM Murray Sargent <murrays@exchange.microsoft.com> wrote:
> >>>
> >>> Very cool stuff. A less difficult way of achieving the pop-up annotations of the first link is to put href’s on the variables. Perhaps that’s what Bruce and David are doing…
> >>>
> >>>
> >>>
> >>> Thanks,
> >>>
> >>> Murray
> >>>
> >>>
> >>>
> >>> From: Deyan Ginev
> >>> Sent: Friday, September 11, 2020 9:25 AM
> >>> To: public-mathml4@w3.org
> >>> Subject: [EXTERNAL] Notable mention of rich math applications, "details-on-demand"
> >>>
> >>>
> >>>
> >>> Dear MathML Refresh enthusiasts,
> >>>
> >>> A very pleasing review article was published at DistillPub today,
> >>> examining the various communication aspects of "interactive articles".
> >>>
> >>> In Figure 8 they have a small SVG showcase of a feature some of us
> >>> here have experimented with in the past via MathML, providing a short
> >>> "legend" of each constituent of a math expression.
> >>>
> >>> Pieces of their example directly overlap with our main scope (names
> >>> for dot product, integral over closed surface), and pieces are in the
> >>> gray zone we are currently discussing (q is "the amount of charge in
> >>> coulombs")
> >>>
> >>> https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdistill.pub%2F2020%2Fcommunicating-with-interactive-articles%2F%23details-math&amp;data=02%7C01%7Cmurrays%40exchange.microsoft.com%7C416ae94ca77c4468495908d8566f4e0a%7C72f988bf86f141af91ab2d7cd011db47%7C0%7C0%7C637354383352852160&amp;sdata=RlIcBzOkmBNSBBrfnQkMNL9xn19tJmnhdVCRBXMCcKc%3D&amp;reserved=0
> >>>
> >>> They also linked to a nice older resource I remember, which seemingly
> >>> introduced the didactic technique of coloring math in web documents
> >>> with colored text in parallel:
> >>>
> >>> https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbetterexplained.com%2Farticles%2Fcolorized-math-equations%2F&amp;data=02%7C01%7Cmurrays%40exchange.microsoft.com%7C416ae94ca77c4468495908d8566f4e0a%7C72f988bf86f141af91ab2d7cd011db47%7C0%7C0%7C637354383352852160&amp;sdata=DPD2kOpiBDb6VGEuhjopZUKevR9U1vBGnPG3%2FcqLnIs%3D&amp;reserved=0
> >>>
> >>> One thing to remark is that this "details-on-demand" application is
> >>> out-of-scope for our "a11y semantics" charter. At the same time, the
> >>> application is - at least in my mind - requiring the same "degree of
> >>> annotation" to be enabled, as the minimal requirements are 1) knowing
> >>> the operator structure/content tree and 2) knowing the
> >>> names/properties of the constituent objects. So this may be a
> >>> "sibling" or "cousin" application to the a11y/information retrieval
> >>> applications we've been discussing.
> >>>
> >>> Something to chew on, and wishing everyone a great weekend!
> >>>
> >>> Greetings,
> >>> Deyan
> >>>
> >>>

Received on Monday, 14 September 2020 12:32:07 UTC