W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > January to March 2010

[TF-ENT] Updated doc with D-Entailment, revised Direct Semantics and proofs

From: Birte Glimm <birte.glimm@comlab.ox.ac.uk>
Date: Fri, 19 Mar 2010 12:49:25 +0100
Message-ID: <492f2b0b1003190449v78e25bacoc8ca674d74d5c3e1@mail.gmail.com>
To: SPARQL Working Group <public-rdf-dawg@w3.org>
Hi all,
I updated several parts of the doc. I added D-entailment (just
following the same line as the others), I describe the mapping from
triples to OWL objects more precisely (Matthew Horridge suggested to
have a grammar and I now outline how we extend the grammar from the
OWL 2 structural spec by allowing for variables and how the OWL
mapping from RDF graphs to OWL objects can be extended to work with
variables). This means that we now have a clear definition of
well-formed queries for the Direct Semantics regime and several of the
conditions are now obsolete since the grammar captures this. OWL
Direct Semantics is also much closer to the other regimes now. As long
as you have well-formed graphs and BGPs, basically the same C1 and C2
apply (just the scoping graph SG becomes the ontology O(SG) obtained
from SG). In addition to the standard C1 and C2 we also have C3, which
requires that adding the instantiated triples to O(SG) still gives an
OWL 2 DL ontology. This prevents, for example, that I instantiate an
object property variables in a cardinality constraint with a
non-simple role. This would make reasoning undecidable and cannot be
captured by a grammar, but is defined in the OWL 2 DL constraints.
This is now all that is required for Direct Semantics, so the regime
is much better in line with the others. There is an optional condition
C4, which can be used to make datatype reasoning feasible (for the
general case we are not yet sure how to do that in all cases). Since
this is optional, implementations could drop it when better algorithms
are know (similar as what OWL 2 does by allowing implementations to
only cover datatypes that can be mapped to system primitives).

I also added an appendix with proofs that show how the regimes satisfy
the conditions on extensions of BGP matching and fixed al broken link
and mark-up problems.

@Chime: I moved one </p> in your section to make the validator happy:
<p>This should result in the following bindings as a result of the
rules and the triples (2)-(7) from a SPARQL service that implements
the RIF Core entailment regime:**</p>** (<- was after the following
table before)

What is no longer in the Direct Semantics regime is the "combined
semantics", which modified OWL 2 Direct Semantics to also take into
account non-logical stuff such as annotations. I much prefer not
changing the entailment relations from other specs. I can see that
users want to query for annotation, but I would prefer to at least
have an additional custom regime that is an extension of the Direct
Semantics regime. Any opinions? Put it back in? Add an additional
extended direct semantics regime? Wait whether we get complaints?

I think the regimes are now in a pretty good shape for the F2F. I am
happy to hear your opinion :-)

Dr. Birte Glimm, Room 306
Computing Laboratory
Parks Road
United Kingdom
+44 (0)1865 283529
Received on Friday, 19 March 2010 11:49:58 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:41 GMT