W3C home > Mailing lists > Public > public-owl-wg@w3.org > April 2008

ISSUE-122 (QCR RDF syntax): the "Expanding" RDF syntax for QCRs will damage OWL 2 Full

From: OWL Working Group Issue Tracker <sysbot+tracker@w3.org>
Date: Mon, 21 Apr 2008 09:24:35 +0000 (GMT)
To: public-owl-wg@w3.org
Message-Id: <20080421092435.74E7EC6DB0@barney.w3.org>

ISSUE-122 (QCR RDF syntax): the "Expanding" RDF syntax for QCRs will damage OWL 2 Full


Raised by: Michael Schneider
On product: 

The current RDF syntax for qualified cardinality restrictions (QCRs) expands the existing RDF syntax for ordinary cardinality restrictions (CRs) by adding an "owl2:onClass" property to it:

  _:x rdf:type owl:Restriction .   # old CR syntax
  _:x owl:onProperty p .           # old CR syntax
  _:x owl:cardinality 1 .          # old CR syntax
  _:x owl2:onClass C .             # new QCR syntax

As originally observed by Peter Patel-Schneider [1], this is a problem for OWL 2 Full, because every QCR is then also an ordinary CR: While the new semantic conditions for QCRs will specify semantics for the whole 4-triple expression above, the already existing semantic conditions for normal CRs will additionally specify semantics only for the first 3 triples. And since a normal CR can be seen as a QCR on class owl:Thing, this means that in many cases the same QCR class will have two /different/ class extensions. This will generally lead to heavy problems.

The following example demonstrates the problem:

  _:x rdf:type owl:Restriction .
  _:x owl:onProperty owl:sameAs .
  _:x owl:cardinality 1 .
  _:x owl2:onClass owl:Nothing .

This QCR expression represents the empty set (class extension of owl:Nothing). But it also represents the whole universe (class extension of owl:Thing), when regarded as an ordinary CR. 

Having this QCR within an OWL 2 Full ontology will make this ontology inconsistent. In particular, having a comprehension principle for QCRs in OWL 2 Full will introduce the above QCR into /every/ OWL 2 Full ontology, i.e. the semantics of OWL 2 Full itself would become inconsistent.

But also without a comprehension principle, the use of QCRs will in many cases make an ontology, which is consistent in OWL 2 DL, inconsistent in OWL 2 Full. Note that the problem is /not/ restricted to cases, where OWL vocabulary URIs are used as the class or the property on which the QCR is put, so forbidding these "syntax reflection" cases would not help.

There are different ways to remove this problem. See [2] for a compact summary and further pointers.

[1] <http://lists.w3.org/Archives/Public/public-owl-wg/2007Dec/0095.html>
[2] <http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0111.html>
Received on Monday, 21 April 2008 09:25:09 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:42:04 UTC