W3C home > Mailing lists > Public > www-rdf-logic@w3.org > January 2001

RE: DAML+OIL: Questions & Improvements.

From: Ian Horrocks <horrocks@cs.man.ac.uk>
Date: Sat, 27 Jan 2001 18:00:42 +0000 (GMT)
Message-ID: <14963.3274.573515.438502@excalibur.cs.man.ac.uk>
To: "Dickinson, Ian J" <Ian_J_Dickinson@hplb.hpl.hp.com>
Cc: "'King . Dany'" <DKing@drc.com>, "'www-rdf-logic@w3.org'" <www-rdf-logic@w3.org>
Hi Ian,

On January 26, Dickinson, Ian J writes:
> Hi Dany,
> Regarding your question 5, it seems to me that you're confusing two things
> (given my limited and still evolving understanding of ontologies and
> DAML+OIL).  Class, property and instance are the modelling primitives
> provided by RDF.  They provide a meta-language for defining an ontology,
> part of which is a hierarchy of concepts.  The most general concept is (in
> DAML+OIL) "Thing".  Every modelled concept is a Thing. The least general
> concept is "Nothing". No modelled concept is a Nothing.  Concepts are
> related by a number of different relations, one of which is sub-class.  If A
> is a sub-class of B, every B is an A, but not vice-versa.  So the bug in the
> example, imho, is that Animal should be a sub-class of Thing.

This sounds about right, but you should be very careful to
differentiate between instance (rdf:Type) and subclass
(rdfs:subClassOf) relationships. When you say "Every modelled concept
is a Thing" you should have said that every modelled daml+oil concept
is subClassOf Thing. Similarly, Nothing is a subClassOf every daml+oil
concept.

> Relations like subClassOf, complementOf, etc, are intended to describe
> relationships among concepts.  Given the modelling meta-language, concepts
> are modelled as RDF classes.  Therefore the rdf:type relation is being used
> to denote precisely that: every concept (Thing, Animal, etc) is modelled as
> rdfs:Class.  But because it's part of the metalanguage, rdfs:Class does not,
> in fact, appear in the concept hierarchy.
>
> As far as I can tell from the various versions, at one time DAML relations
> like type were considered distinct from RDFS and RDF relations. Then in the
> 2000/12 version, where there were overlaps in terminology, the RDF or RDFS
> relations were used in preference. It seems to me that you're pointing out a
> potential problem with this, which is:
>    Thing rdf:type rdfs:Class
> means "concept Thing is modelled as an RDFS Class", whereas
>    adam rdf:type Person
> means "instance adam belongs to the concept Person".

Remember that the elements of rdfs:Class can themselves be
classes. daml+oil classes (like Thing and Person) are just such
instances of rdfs:Class. adam is an instance of the class Person.

Ian
Received on Saturday, 27 January 2001 13:13:38 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:38 GMT