RE: comment: WD 10 May 2005

Bernard,
Thank you for the reply. I don't deny the need for a class like Concept. I guess
that I'm disturbed by your implementation. One use case I think of categorizes
resources by the terms defined within an ontology. Under SKOS, the whole
ontology seems duplicated so that each resource's <dc:Subject> value can
reference an instance of Concept rather than directly referencing a class
defined in the ontology.

CONCEPT DEFINITION. What is "a term in a controlled vocabulary" (your words) if
not an rdfs:Class -- this is not asked rhetorically at all. Perhaps your
definition of a Concept is revealed when you say with regard to FreedomOfSpeech
that "I am quite reluctant to consider this concept as a class, because I wonder
what the instances would be" -- this suggest your definition is that anything
that is difficult to envision as an instance, is therefore a Concept. [BTW,
consider a "FreedomOfSpeech" resource as an item in an inventory of rights had
by an individual.]

My alternative view is that a 'concept' is a document-related class of resources
distinguishable by a Reasoner from the classes of terms defined within an
ontology being used by or specified within the document. Resources can be
categorized as being 'of' a concept either by a property, eg <dc:Subject>, or by
an <rdf:type> element which specifies the Concept instance. Conversely, the
'isSubjectOf' property should most definitely be applicable to instances of
owl:Class (and to its subclass, Concept), referring to those resources
categorized as being 'of' the concept.

TOPIC RELATION. I see no conceptual semantic difference, and little has been
offered, between a concept and a topic. My view is that every concept IS a
topic; every topic is a concept. A topic (or concept) refers conceptually either
to (a) a set of generic or specific instances or (b) a specific instance, about
which statements are made or, said another way, which are the subject matter of
ensuing statements. Due to (a), I claim that topics are normally plural terms;
those infrequent times they are singular, means that the topic of the text unit
is a specific resource.

An ontology class can be referenced in a ConceptSchema as a Concept that is
either a reified statement about the class, or one whose subject is the ontology
class. In this way class instances can be referenced in a topic map as easily as
new topics (or concepts) can be defined. Likewise, a topic or concept that is
about another topic or concept, is a reified statement or (if you prefer) one
whose subject is the other topic or concept.

Lastly, and perhaps this is better left for later, I see little difference
between a 'category' and a concept, topic, or class -- in my own ontology I use
the metaclass 'Category' because it seems more user-friendly. Its Word Net
definition is "A general concept that marks divisions or coordinations in a
conceptual scheme. A collection of things sharing a common attribute."

OTHER SKOS CLASSES. I see no problem making skos:ConceptSchema (or TopicMap) a
subclass of Ontology. A skos:Collection looks all the world like an anonymous
class (which can have its own rdfs:label BTW) and which is composed of a unionOf
(with its parseType =Collection incidentally) with its sub-concepts (or
sub-topics or sub-classes) then listed. Can OrderedCollection be accommodated by
a different parseType?

SKOS PROPERTIES. If a Concept is a metaclass (that is, whose instances are also
instances of owl:Class), there are implications for the properties defined for
Concept. For instance, while the 'broader' property appears directly equivalent
to rdfs:subClassOf, the 'narrower' property is not formally accommodated. Child
concept/topic resources certainly are narrowed concepts/topics in accordance to
the definition of a 'subclass'.

I haven't time now to analyze other SKOS properties... Bottom-line, the SKOS
data model remains ambiguous to me so I would much appreciate it if SKOS could
make a rigorous distinction between a class, topic, concept, subject, and
category -- to help practitioners like myself understand best practices on the
Semantic Web.

Oh, as for "table of contents" -- I can't find it now! Thought it was there....
sorry for the noise about that.

Best regards,
John McClure

Received on Thursday, 21 July 2005 00:11:06 UTC