Re: What's the intent for ":"?

Using a literal on a leaf (to get desired speech) as opposed to a concept
name is a good point, one that I don't think I'm alone in forgetting. I did
a quick scan of the spec and it seems we don't specifically call that out.
I think an additional example would be particularly useful.

Your reminder that it should be a literal solves the problem for me. People
should probably use
   <mo intent='ratio'>:</mo>
and the AT should know what to speak (assuming "ratio" is a core concept).

I'm less sure "time" is an appropriate core concept name as I suspect it
could encompass dates and other time. Maybe that's ok, or maybe it's too
generic. I'd need to see a list of all the things it encompasses to know
whether it introduces ambiguity or not.

Although I can see the appeal of using an intent value of "_to:ratio", that
means almost all concepts can be properties. But concepts have arity and
airity doesn't exist on tokens. So a concept name and a property of the
same name differ. It seems like a can of worms. Of course, there is the can
of worms of why "ratio" on a leaf is ok, but assumedly (by me),
intent="transpose" (as opposed to "_transpose") on a "T" is not ok. The
devil is in the details...

    Neil


On Thu, Jul 27, 2023 at 7:06 PM Deyan Ginev <deyan.ginev@gmail.com> wrote:

> Hi Neil,
>
> I'm focusing on the single "ratio" in the latter half of your email. As
> you say, the legal concept annotation is intent="ratio".
> Selecting explicit English prepositions should be done via a literal, i.e.
> intent="_to" or intent="_for". Similarly the empty literal override for
> time would be intent="_".
> In those cases we could enforce speech, but will lose the concept
> information.
>
> Similarly for Bulgarian prepositions: one can choose between the literals
> intent="_към" or intent="_за", but the concept is still intent="ratio",
> which is localized intent="съотношение", with a possible alias name
> "отношение".
>
> I think being disciplined about the distinction between (mathematical)
> concepts and (language) literals is one of the lifelines we have available
> in making these decisions, while retaining a clear naming scheme.
>
> Some additions:
> 1. The wikipedia page on ratio[1] suggests that one can also encounter the
> notation "a to b" directly, where one would assume an <mtext>to</mtext> in
> MathML, possibly with wrapping spaces.  So a Braille reader may also
> encounter the "to" word directly as text.
>
> 2. Wikipedia also provides an alternative readout of a ratio when the
> context is a "proportion", i.e. an equality between two ratios. In "a:b =
> c:d" the left-hand side "a:b" can be narrated "a is to b", where the "c:d"
> right-hand side can also be narrated "c is to d", with the connecting "="
> read "as".
>
> In an AT that wants to provide specialized readings for proportion, one
> could imagine recognizing a wrapping mrow property <mrow
> intent=":proportion">.
> Or alternatively, having a pattern that recognizes infix use of "=" where
> both arguments can be identified as ratios.
> On the argument level, AT can either try to detect the infix use of <mo
> intent="ratio">:</mo>, or cases where the argument is wrapped with a
> property, as in <mrow intent=":ratio">.
> The wrapping mrow could be a helpful technique for something harder to
> detect such as the n-ary ratio "2 : 4 : 8 : 16".
>
> I think embracing the "progressive enhancement" spirit could be healthy in
> this kind of ladder of ever increasing annotation burden. Does the
> completely raw MathML get narrated well? If yes, leave as-is. If not,
> annotate the innermost nodes that encounter issues. If the speech remains
> unnatural at places, gradually climb up the subtrees, depositing properties
> and compound intent expressions, until the desired outcome is achieved.
>
> Greetings,
> Deyan
>
> [1] https://en.wikipedia.org/wiki/Ratio
>
> On Thu, Jul 27, 2023 at 9:30 PM Neil Soiffer <soiffer@alum.mit.edu> wrote:
>
>> I think everyone would agree that ":" is ambiguous. It's one of the few
>> cases where the braille Nemeth code (but not UEB) cares about the meaning.
>>
>> Here's a case I came across:  3:30
>>
>> It certainly could be a ratio. Or it could be a time. In normal use,
>> there might be more context to help (3:30 = 1:10 or 3:30pm). However, when
>> standing alone, there is no way to know.
>>
>> I think both ratio and time are common enough to be in core. Probably
>> also function definition (f: x → y) is in core along with "such that"
>> ({x: x>0}).
>>
>> As per the discussion today (and times in the past), the intent should be
>> on ":" when possible. However, ":" is a case where it doesn't work out
>> well. One can force the speech (e.g., "intent='to' for ratio or intent=' '
>> for time). But how does the braille know that 'to' means it is a ratio and
>> should output the braille for ratios instead of the braille for a
>> punctuation ":"?
>>
>> Aside: listening to Khan Academy, "to" is used in something like "3:2"
>> but "for" or "for every" is used in the videos when there are
>> quantities/units given as in "3 dogs : 2 cats". So here you would use
>> intent="for" -- so saying AT should know that "to" used on ":" means ratio
>> is not going to work.
>>
>> I think this is a case where pushing intent to the operator doesn't work
>> despite it being an infix operator and it needs to be on the mrow. Being in
>> core, an intent value of "ratio" could be spoken as "to" by AT, but still
>> recognized easily when generating braille.
>>
>> Thoughts?
>>
>>     Neil
>>
>>