[topicmapmail] Are Facets Really Simple After All?

Thomas B. Passin tpassin@comcast.net
Mon, 01 Dec 2003 20:30:44 -0500


Kal Ahmed wrote:

> However if I were to transcode property/value pairs which are clearly
> properties of subjects described by a topic in my topic map, I would use
> associations for those property/value pairs where the value is a
> resource which is the subject or object of other statements, and
> occurrences for those property/value pairs where the value is a resource
> that does not appear in any other statement or where the value is a
> literal string. But thats just a rule of thumb.
> 

In one way of looking at things, properties are named values that relate 
to a single subject, whereas associations relate two (or more) subjects. 
   With this view, you would be very interested in using a different 
mechanism for properties and associations, because you see them as 
fundamentally different kinds of things.

 From another point of view, properties and associations are both 
predicates.  The first kind has one argument (besides the subject), the 
second has two or more.  In this view, there is no essential difference 
between properties and associations.

I usually prefer the first way, but it depends on what you want to to. 
Here are three conceptual graphs from p. 31 of John Sowa's book 
Knowledge Representation (relations are in parens, concepts are in 
square brackets) -

1) (Red) -> [Ball]
2) [Ball] -> (Attr) -> [Red]
3) [Ball] -> (Chrc) -> [color:Red]

Here, "Chrc" is short for "characteristic". and "Attr" is short for 
"Attribute".

It seems to me that we cannot quite do either 2)  with topic maps, 
because we cannot use a topic as the value of an occurrence.

We could do 3) as an occurrence, I think, by creating a topic "color" 
and giving an instance of it an occurrence whose type would be "value" 
and whose value would be "Red".  The "Chrc" relation would be implied by 
the use of the occurrence.

We can do 2) if we create a topic for "Red".

Hmm... anybody want to comment?

Cheers,

Tom P