W3C home > Mailing lists > Public > www-webont-wg@w3.org > April 2003

RE: Suggestions for response to: Some comments on OWL Reference

From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
Date: Thu, 10 Apr 2003 17:33:49 +0200
To: "Guus Schreiber" <schreiber@cs.vu.nl>, "WebOnt WG" <www-webont-wg@w3.org>
Message-ID: <BHEGLCKMOHGLGNOKPGHDAEHLCBAA.jjc@hpl.hp.com>

>  >
>  > It seems to me that:
>  > 1) There is no PRECISE SYNTAX  of OWL.
>  > 2) The XML encoding of an OWL ontology is based on RDF/XML Syntax as
> well as RDF schema for OWL (Appendix B).
>  > 3) Without the PRECISE SYNTAX of OWL, where does the *Final and
> Formally* stated normative definition of the language come from?
>  >
>  > I think that the specs should give a consistent and explicit stating
> on this issue.
>  >
>
> I suggest we discuss this at the telecon.
> I'm not sure about the response. It is true that Appendix B is the basis
> for the RDF/XML syntax, so dropping its normative status has some
> drawbacks. From S&AS the RDF/XML syntax can only be indireclty derived.
>
> A proposal could be to restore the normative status of the Appendix B.
>


There are wording issues - Qu might come back with a substantive objection
if he understands what we are trying to do.

The abstract syntax is informal - but the mapping rules into RDF/XML are
sufficiently formal to address (3). This approach is sufficiently obscure
that it is conceivable that there is not enough text making it clear.

It perhaps would have helped if we had got the triple table version in as
well.


>
>
>  > 2. The domain and range of owl:equivalentClass in OWL Lite.
>  >
>  > [    3.2.2 owl:equivalentClass
>  >     ...
>  >     NOTE: OWL DL does not put any constraints on the types of class
> descriptions that can be used as domain and range values of an
> owl:equivalentClass statement. In OWL Lite **only class identifiers and
> property restrictions** are allowed as domain and range values. (?)
>  >
>  >     8.3 OWL Lite
>  >     ...
>  >     the subject of owl:equivalentClass triples be named classes and
> the object of owl:equivalentClass triples be named classes,
> restrictions, or subjects of owl:intersectionOf triples (?);
>  >     ...
>  > ]
>  >
>  > 1) According to S&AS, the domain of owl:equivalentClass must be just
> classID.
>  >
>  > 2) As to the range of owl:equivalentClass, class identifiers and
> property restrictions are certainly allowed as range values. But how
> about others allowed as range values? What's "the subjects of
> owl:intersectionOf triples" mentioned in section 8.3?
>  >
>  > It seems most likely to be anonymous classes defined as the
> conjunctions of class identifiers and property restrictions.
>  >
>  > It (The domain and range of owl:equivalentClass in OWL Lite) should
> be explicitly and consistently specified.
>  >
>
> I suggest we discuss this at the telecon.
>
> We changed this at a late point, but I cannot find the record. I seem to
> remember we opted for allowing only class IDs in equivalent classes for
> OWL Lite. This is also what the OWL Lite class axioms in S&AS state.
>
> Proposal: to make editorial changes to Ref (3.2 and 8.3) and S&AS such
> that only class identifiers are allowed for equivalent classes in
> OWL Lite.
>
>

My understanding is that
  In OWL Lite
    owl:equivalentClass has subjects being classID
              and object being classID or blank restriction nodes.
  In OWL DL
    owl:equivalentClass has subjects and objects both being arbitrary
description nodes (classID or blank nodes of type owl:Class or
owl:Restriction); subject to the constraint that there must be a Hamiltonian
Path through at least one of each of the following sets of directed
sub-graphs:

From an RDF Graph G form a set of subgraphs S

S = { x | x subgraph of G, all nodes in x are blank,
          no owl:equivalentClass edge links (in either direction)
          any node in x to any blank node not in x   }

i.e. S is the set of maximal owl:equivalentClas-connected blank sub-graphs
of G. Note the empty graph is always in S.

Then if G is in OWL DL the syntax of owl:equivalentClass then there exisits
a set T of subgraphs of G such that for every member t of T:
- every edge in t is labelled owl:equivalentClass
- t has a directed Hamiltonian path

and moreover for each s in S there is a t in T such that s subgraph of t.

And yes, that is a mistake.

Jeremy
Received on Thursday, 10 April 2003 11:34:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:58 GMT