RDF MT: RDFS entailment issues

Dear all,

We have recently been implementing a new inferencer for Sesame,
that uses the RDFS entailment rules as specified in the MT
working draft.  While testing this, we came across two issues
with the current set of rules.

In rule #8, it is stated that any class is a subclass of
Resource.  Applying this rule in practice resulted in
rdfs:Literal also being defined as a subclass of Resource. We are
aware that this is an open issue within the RDF Core WG. The
current spec still defines Literals as disjoint from Resources,
however. Is the MT assuming the issue will be resolved in this
fashion, or has this simply been overlooked?

This issue has also been raised in the "Compatibility Guide" by
Wolfram Conen and Reinhold Klapsing[1], but we thought it would
be a good idea to point it out again, to stress that this is
encountered in practical applications.

Another issue is that of reflexivity of the subClassOf relation.
The RDF Core WG has resolved that cycles are allowed in
subClassOf hierarchies.  Reflexivity of the relation seems a
natural consequence of this, since it is in a sense the most
trivial form of a cycle.

Nevertheless, this is not captured in the current entailment
rules.  Would it be a good idea to add a rule #8a:

   if E contains:

    xxx rdf:type rdfs:Class

   then add:

    xxx rdfs:subClassOf xxx

to make the schema closure complete (and a similar rule for
subPropertyOf, of course)? This is compatible with the rest of
the MT since it is explicitly stated that the subset relation
need not be proper (i.e. the subset can be equivalent).


Best regards,

Jeen Broekstra and Arjohn Kampman

[1] http://nestroy.wi-inf.uni-essen.de/rdf/new_interpretation/RDFS-LogInt_v10.pdf
-- 
jeen.broekstra@aidministrator.nl
aidministrator nederland bv - http://www.aidministrator.nl/
julianaplein 14b, 3817 cs amersfoort, the netherlands
tel. +31-(0)33-4659987, fax. +31-(0)33-4659987

Received on Friday, 1 February 2002 06:55:41 UTC