Re: RDF-ISSUE-79 (undefined-datatype): What is the value of a literal whose datatype IRI is not a datatype? [RDF Concepts]

Hi Antoine,

On 15 Nov 2011, at 21:23, Antoine Zimmermann wrote:
>> if<bar>  owl:sameAs<baz>, and<baz>  is an IRI in the datatype map,
>> then "foo"^^<bar>  may have a well-defined value even if the IRI<bar>
>> is not in the datatype map.
> Assuming that owl:sameAs was in the RDF/RDFS spec, this even would not
> make "foo"^^<bar> be interpreted identically as "foo"^^<baz>. The interpretation of typed literals is not influenced by anything in the ontology, even in OWL. "foo"^^<bar> is always interpreted as L2V(D(<bar>))("foo") according to the datatype map D.
> FYI, look at section 4.2 of the OWL 2 RDF-based semantics:
> "IL is a mapping from typed literals "s"^^u in V to their denotations in IR, where IL("s"^^u) = L2V(d)(s), provided that d is a datatype of D, IS(u) = d, and s is in the lexical space LS(d); otherwise IL("s"^^u) is not in LV."

That just requires that *d* (the datatype) is in the datatype map, not that *u* (the IRI) is in the datatype map. If I have <u> owl:sameAs <v>, then it follows that IS(u)=IS(v). If (v,d) is in the datatype map, then it follows that IS(v)=d, which gives us IS(u)=d without u being in the datatype map.

That's how I read it anyways.

> So the interpretation of "s"^^u (or "foo"^^<bar>) is in the value space of <bar> if and only if <bar> is in the datatype map

If my reasoning above is correct, then it doesn't require that <bar> is in the datatype map. It requires that <bar> denotes d, and that d is in the datatype map.

> Otherwise it is not in LV 

(That's actually not a requirement that I can find anywhere in RDF Semantics. RDF Semantics says that, if a literal's datatype IRI doesn't denote a datatype, then the literal “is treated as before”, it denotes “some unknown thing”. In other words, its denotation is entirely unconstrained, and interpretations that map the literal to a member of LV are fine. So the wording you quote looks a bit like a bug in OWL2 to me.)

> I'm just saying that the word "value" may be misunderstood. For instance, I would not consider a person to be a value.

I don't think there's any basis for that. There's nothing *formally* wrong with defining a datatype whose lexical space is social security numbers, whose value space is citizens, and whose L2V mapping involves the social security database.


Received on Wednesday, 16 November 2011 21:14:05 UTC