[sc34wg3] Association items

Murray Altheim sc34wg3@isotopicmaps.org
Sun, 12 Jun 2005 15:52:59 +0100

Lars Marius Garshol wrote:
> * Murray Altheim
> | 
> | Actually, "unary name" and "unary occurrence" do sound a bit
> | strange, but are accurate descriptions from an ontological modeling
> | perspective.
> Possibly, but you can't have them in topic maps. They would equate to
> a name that is not the name of any topic, or an occurrence that is not
> the occurrence of any topic. Neither really makes any sense from a
> topic map point of view.

In reading over your reply, I think you and I are perhaps using
terminology differently, or understanding terminology differently.

"unary name" and "unary occurrence" are to my understanding (though
I'd not term it this way) the way things are already done within a
Topic Map. If you accept the notion (ref Sowa) that a monadic
predicate or unary association (basically the same thing, different
terminology) is a fancy name for a property, and you consider a
property as a synonym for a characteristic, then a Topic characteristic
(of which name and occurrence are two instances) is related to the
Topic via a unary association.

[if that still doesn't make sense or we still disagree, let's then
step back a bit and try again...]

> | Unary associations are also called monadic predicates, which are by
> | some called "properties" (e.g., Sowa's "Knowledge Representation",
> | p469).  Topic names and occurrences are what we call "Topic
> | characteristics" but what in the KR/OntEngr field would often call
> | properties of the Topic. They are related to the Topic via a monadic
> | predicate. 
> How can you relate one thing (a name or an occurrence) to another
> thing (a topic) with a unary association? Or are the terms "unary
> associations" and "monadic predicate" not equivalent?

In the grand scheme of entity relations, not everything considered an
entity in the model. So if you're in, say, Protege, you don't want to
have to add a new entity into the model if you just want to name an
existing entity. So you give that existing entity a name, which is a
kind of property. The relationship between the entity and its name is
via a monadic predicate or unary association -- same basic thing --
though "monadic predicate" is to be precise an indicator of the number
of arguments in the relation. As Sowa says: "Some authors say that
relations must have two or more arguments; they would call a predicate
with one argument a property. However, the terms predicate and relation
are often used interchangeably." The term "unary association" is meant
to refer to a relation with one argument, though my own choice of using
"association" is meant to refer to the domain of Topic Maps. I generally
use "relation" when speaking about relations in general. (And I generally
capitalize Topic and Association to be more obvious about this.)

> | Had we decided not to provide the syntactic sugar in XTM, they could
> | have been modeled using typed <association> elements.
> Provided <association> was extended somewhat.

No, by this I meant that if we hadn't provided element types like
<basename> and <occurrence>, this could have been done via typed
Associations. One could take an XTM 1.0 document and decompose it
by breaking each Topic characteristic to its equivalent as an
Association between the characteristic and its Topic, but that sugar
has good purpose in simplifying the markup.

By the same token, it would be possible to take an XTM 1.0 document
and represent *everything* in it via GXL, by using GXL's edges typed
as Associations to connect nodes typed as Topics. It would be quite
a lot more verbose, but I see no real barrier technically.

> * Lars Marius Garshol
> |
> | You mean that it would be equivalent to
> |   is-inquisitive(jan : person, is-inquisitive : characteristic)?
> | That seems like a bizarre interpretation to me. Why would we do this?
> * Murray Altheim
> |
> | Well, that's precisely what we did in ISO 13250 and XTM 1.0. We have
> | been calling these things "Topic characteristics" and the relation
> | they have with the Topic is a monadic, property relation. As Patrick
> | points out, the role is basically characteristic/entity or character-
> | istic/Topic (in our case).
> I agree with your description of topic characteristics, but I don't
> see how that equates to the suggested interpretation I gave above for
> unary associations.

Perhaps we've been using (as I suggest above) our terms differently.

Does this help? Get us closer to agreement or understanding?


Murray Altheim                          http://www.altheim.com/murray/
Strategic and Services Development
The Open University Library
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .

     [Tom Cruise] has attained an Operating Thetan 6, a high
     level of purity (there are eight levels) which means he
     knows the space alien story called 'the incident'. He
     recently got into a squabble with a journalist for Der
     Spiegel who called Scientology a pseudo-science.

     OT III: http://www.xs4all.nl/~kspaink/fishman/ot3.html
     Operation Clambake: http://www.xenu.net/