Re: please read before our meeting...

Hi Paul,

Why do you suggest "known" / "uknown" are weak names? To me they are very
close to the healthy balance the spec needs:

1. They are needed, as we recommend behaviors for each of the two cases

2. They avoid specificity of data structures, implementation, even AT
paradigm, which allows ample room for experimentation.

3. They also help indicate that the spec is focused on the Intent syntax,
and not on the technical details of AT systems, which makes the text easier
to read confusion-free.

I quite like them personally.

Greetings,
Deyan

On Fri, Oct 25, 2024, 4:29 PM Paul Libbrecht <paul@hoplahup.net> wrote:

> While ‘known/unknown’ appear weak as names, I really really liked the idea
> that an AT considers an extra “intent concept/property dictionary” as a
> declarative way to indicate what extra language or specific culture they
> consider. This is a first and very basic step towards
> context-specific-pronunciation. Maybe ‘considered/not-considered’, or
> ‘in-scope/out-of-scope’ ?
>
> Btw, the text I read used “AT” without a determinant (“An AT”/“The AT”).
> That sounded very odd as language for me (I hear this from particular
> immigrants here ;-)). That was just a speedy writing effect, right?
>
> Paul
>
>
>
> On 24 Oct 2024, at 16:49, Neil Soiffer wrote:
>
> I mostly agree with Bruce's suggestions. Here's my take:
>
> 5.1: I think known/unknown concepts should be dropped and text added along
> the lines of "See 5.2 for more information about how concepts are used".
>
> 5.2: I think the flow is:
> 1. Define the Core concept list followed by the open concept list.
> 2. Define the Intent Concept Dictionary (should use core, may use open,
> may have other entries)
> 3. Describe when an intent matches an entry in the dictionary and define
> known and unknown concepts.
> 4. Describe what AT should do when there is a match and when there isn't a
> match.
>
> 5.3: Should continue to be about properties and include what is in 5.4
> (what to do if only properties are given)
>
> I'm not sure about "5.4 (New!) How to apply Concept & Property Hints".
> That is sort of what is in '5.7 Intent Examples". That section is
> non-normative (do we need to state that?) and maybe that is different from
> what Bruce is suggesting.
>
>     Neil
>
>
> On Wed, Oct 23, 2024 at 12:04 PM Bruce Miller <bruce.miller@nist.gov>
> wrote:
>
>> Hello again;
>>
>>    So, my biggest problem with this section is that it tends to be too
>> distributed; as you keep reading, you keep accumulating clarifications,
>> corrections, nuance, so that you don't really know what eg. "known" is,
>> or how to use it, or if it's even consistent, until you get to the end &
>> you've assembled all the pieces.  This may make it nice to read as
>> literature, but hard to use as a specification.
>>
>> My suggestions (with lots of hand-waving):
>>
>> 5.1 Grammar for intent: should focus on the grammar and it's
>> terminology, but not get into how it's used.  So under the "concept"
>> item, everything after "A known concept..." should be pushed back to
>> 5.2. OR at most replaced by "A concept may be known or not, see 5.2".
>>
>> 5.2 Intent Concept Dictionaries: should focus on describing the
>> dictionaries, and how concepts are matched (and thus should define
>> Known/Unknown), but still should defer how the entries are used. So,
>> under item Core, all but the 1st paragraph should be pushed back to (a
>> new) 5.4.
>>
>> 5.4 Intent Self References: doesn't seem to warrant it's own section.
>> Can't it be stated in 5.3 that a property can stand alone, w/o a concept?
>>
>> 5.4 (New!) How to apply Concept & Property Hints: This should collect in
>> one place how known concepts, unknown concepts, literals,  might be
>> spoken, with whatever level of compulsion, and how properties may or may
>> not modify them.  If we have it in one place, any contradictions may be
>> easier to detect :>
>>
>> Aside: I have a tendency to think of "Concept" and "Property" as
>> corresponding to "What" and "How", but this projection isn't completely
>> consistent with all our use cases, or terminology. Should it be? I
>> dunno, but at least that may explain some of my prejudices :>
>>
>> bruce
>>
>