RE: Close to final draft of "classes as values" note

Brian

> Approach 2 confuses me.  It seems to say that a LionSubject is of type lion,
> which I interpret to mean it's the sort of individual with claws that I
> don't want to meet close up when its hungry.  It doesn't seem intuitive to
> me that the book "Lions: Life in the Pride" is about a particular individual
> lion.  It seems to (an inexperienced and naïve) me that the relation between
> LionSubject and Lion should not be rdf:type.  Again, it seems to me that the
> diagram of approach 2 is in RDFS.  Doesn't need Owl DL.

Agreed. see my previous post about it.

> Approach 3:  There is presumably a relation between LionSubject and Lion.
> Do we have any vocabulary to describe it?  Here we really need Owl to define
> the transitive nature of the thesarus relations.

I think the relationship, or indeed the *gap* between "LionSubject" and "Lion" is really
the core issue. Approach 3 is the closest to real world situations. The right side of the
diagram is the library world, with its classification and indexing schemes. There you have
documents, classes of documents, and indexing schemes. "LionSubject" would certainly be
better called a "Lions" category in such a scheme. See my previous post, and e.g.
http://isbn.nu/sisbn/lions/.

The ontology world at the left is completely disconnected from the libray world. So the
real question is : how are those worlds connected?

The final comment about approach 3 is : "We need to maintain a set of instances for all
subjects in addition to the hierarchy of subjects and ensure that the two sets — classes
representing subjects and corresponding individuals and values for the parentSubject
property — are consistent with each other."

But how can one check this "consistency" if there is no relation(ship) between the two
worlds?

My view, already expressed in another thread in answer to Alistair about SKOS and WordNet,
is that the subject or concept "Lions" as in "Do you have any books about Lions?" is the
most generic and interoperable, in the sense that it does not presume of any specific KOS
where the concept would be represented by a formal object with formal properties. The
class "Lion" in a formal ontology, is a specific formalization of the general, informal
concept "Lions", the dmoz or yahoo categories are other formalization of the same concept
in other schemes.
That's what we want to capture. Either we invent a new relation, or we try to see if
available tools would fit more or less.

In the following, a: is the OWL namespace at the left, b: the library scheme namespace at
the right.

A minimalist approach would be :

	a:Lion  rdfs:seeAlso 	b:Lions

A more controversial one would be to acknowledge that the concept is generic to the class

	a:Lion  rdfs:isDefinedBy   b:Lions

And, why not, say that the concept is the subject of the class

	a:Lion  dc:subject   b:Lions

This last one would extend the domain of dc:subject far off its original shores, and would
hardly be approved easily by DC folks, I'm afraid. So we could try a subject indicator à
la topic map.

	a:Lion  tm:subjectIndicator   b:Lions

In any case, the relation would be expressed as an annotation property of the class.

Does that make sense?

Bernard Vatant
Senior Consultant
Knowledge Engineering
Mondeca - www.mondeca.com
bernard.vatant@mondeca.com

Received on Thursday, 13 May 2004 12:47:15 UTC