[topicmapmail] Two Models of Facets

[Murray Altheim]

Peter P. Jones wrote:
[...]
>>Peter,
>>
>>You've thrown a word in that is pretty overloaded around here, so I'd
>>have to ask to have you clarify what you mean before I answer. I don't
>>really know what "template" means in this instance.
> 
> True. OK, I'll try.
> (This is just my take on things, OK - I'm not claiming any 
> righteousness here.)
> I'm talking about an inheritance approach.
> I think it is possible, when thinking about semantic graphs, to ditch 
> the strictness of the class-instance dichotomy at many levels of the 
> graph. 

Ditching strictness just gives you a less strict model. I don't know
if that's gaining anything -- it sounds like you're losing something,
and that you'll likely be adding either ambiguity or arbitrariness.

> For example, in the hierarchy
> Thing -> Animal -> Reptile -> Alligator -> "Allie"
> I believe it is always possible to treat any of those concepts as 
> ones that I want to say a specific thing about. Take 'Animal'. I 
> might say specifically that animals have plasma circulation or a 
> sexual reproduction capability or sensory systems. These are not 
> property-values that I need to explicitly 'pass down' to instances or 
> classes lower in the hierarchy because the value of the property is 
> already satisfied at the level of the Animal class. It isn't a facet 
> in that sense, it's a satisfied property of the Animal topic. 

I don't know that we've defined facet and property well enough to
state such distinctions, but if you're following the definitions
I've set out, I don't see that you've differentiated them at all.
But since you're stating that the Animal topic (which is a class)
has a property, you're not using my definitions, so I don't know
what you define as a facet or a property, or what the differences
you are making between them. So I don't know how to comment accurately
on what you've said, really. You're certainly free to define things
yourself and then describe things in your own terms. I just can't
tell what you're saying without guessing as to your meaning, which
doesn't really lead us anywhere. At least quickly.

> But all animals are deemed to have lifespans or weights too. Take 
> lifespan. I might assign the property 'lifespan' to 'Animal' but give 
> it a value that I know is to be satisfied of constrained further down 
> in the hierarchy of the graph. E.g. "200 > x > 0: [datatype:years]".
> Reptile would inherit that because it is unsatisfied at that level, 
> and might add a further constraint. 

 From what I can tell, all you're doing is putting characteristics
in different places than they'd normally go, which is as high up the
inheritance as is "correct". E.g., while "lifespan" can be assigned
to "Allie", in your hierarchy it'd be correct on "Animal". Your model
is really incorrect otherwise, or at least arbitrary. If a superclass
also has the property, then it should be moved up to the superclass.

> E.g. "110 > x > 0: [datatype:years]". And so on down to Allie, where 
> x gets filled when she's dead.

No, there you're making a mistake semantically, which is exactly the
kind of mistake I'm trying to avoid by keeping "facet" and "property"
distinct: i.e., the "lifespan" of a class is a statistic stating the
average lifespan of a species, whereas Allie's lifespan is literally
how long she as an individual lived. It's these kinds of semantic
mistakes I'm assiduously trying to keep people from making by building
into the model and the language enough differentiation that it's clear'
which is which. The best defense is clear language.

>>My first guess is that a single property-value model perhaps exists
>>across all classes, and that a *different* property-value model exists
>>for individuals (which we've been calling "instances", which is not
>>really correct). I assume you'd agree that characteristics of a class
>>are of a different order than characteristics of individuals? That
>>there is some inherent difference in both the thing being related and
>>the kind of relation? I think there is. E.g., there's a difference
>>between the abstract concept of "height" and the specific height of an
>>individual. The former can't be measured, the latter is a measurement.
> 
> Yes, I've attempted to cater for that above. But maybe it falls short 
> in some respect...?

Well, I'm still unclear on "template". And I don't see that relaxing
the rules between superclass-subclass and class-instance (i.e.,
treating them the same) can ever be considered anything than something
that enables a lot of inferential mistakes to occur.

Murray

......................................................................
Murray Altheim                    http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .

    International lawyers and anti-war campaigners reacted with
    astonishment yesterday after the influential Pentagon hawk
    Richard Perle conceded that the invasion of Iraq had been illegal.
    http://www.guardian.co.uk/uk_news/story/0,3604,1089042,00.html

      "...if it's the case that international law doesn't permit
       unilateral pre-emptive action without the authority of the
       UN, then the defect is in international law."

    Thank God for the death of the UN, by Richard Perle
    http://www.guardian.co.uk/Iraq/Story/0,2763,918812,00.html