RDF Review

Dear All,

Jeff and I (mostly Jeff) have looked at the latest RDF MT, in
particular the Datatypes section. We did not have time for an
exhaustive review, but here are some comments:

Regards, Ian
========================================

1. Interpretations

IP no longer (explicitly) a subset of IR (see defn of RDF simple
interp in Sec 1.3)? Not sure if this has any impact on OWL (or on RDF
come to that).


2. Datatypes

My main impression is that the "datatype clash" in RDF has not
been defined/explained clearly enough in section 5, which presents the
datatype interpretation of RDF. Datatype clash is important
because it is one of *only three inconsistencies* recognized by the
model theory (see section 5). Until the picture is clear, it is
difficult to determine how it affects OWL.

When explaining datatype clash, the semantic doc says

"If the datatypes in the datatype map D impose disjointness conditions
on their value spaces, it is possible for an RDF graph to have no
D-interpretation which satisfies it."

However, it is also possible that the problem comes from the lexical
form, e.g.

<ex:a> <ex:b> "2.5"^^xsd:decimal  (1)
<ex:b> rdfs:range xsd:integer  (2)

Is this a datatype clash?

The only example in section 5 about datatype clash is about typed
literals, what about the cases that we don't use typed literal. E.g.

<ex:c> rdfs:range xsd:string    (3)
<ex:c> rdfs:range xsd:integer    (4)
<ex:d> <ex:c> _:xxx    (5)

Is this a datatype clash? According to the informative entailment
rules in section 7, it is a datatype clash:

from (3) and (rdfs3) we have _:xxx rdf:type xsd:string     (6)
from (4) and (rdfs3) we have _:xxx rdf:type xsd:integer    (7).

If this is the case, does it mean that the informative part of the
document implies something that isn't mentioned in the normative
semantics?

The above example can be regarded as property inconsistency; similarly
we can have class inconsistency:

<ex:e> rdfs:subClassOf xsd:integer    (8)
<ex:e> rdfs:subClassOf xsd:string    (9)

means that ex:e is equivalent to owl:Nothing, and adding

_:yyy rdf:type <ex:e>    (10)

leads to ontology inconsistency.

Received on Thursday, 2 October 2003 12:03:48 UTC