RE: ACTION-186: Review on DL-Semantics

Hello Michael,

Thanks a lot for your review! Below are my responses. Please let me know should you have further comments.

Regards,

	Boris

> The Abstract above sais: "This document provides /a/ model-theoretic semantics ...", while the
> first sentence of the intro talks about "/the/ model-theoretic semantics of OWL 2". I suggest to
> change the text in the /Abstract/ to "/the/ model-theoretic semantics".

Agreed.

> I wonder whether it makes sense to also state at this position that this document provides a
> semantics for the /profiles/.

I added a sentence about this.

> In the definition of "vocabulary" below, a synonym "signature" is given, too. I suggest to drop it,
> since it is never used elsewhere, AFAICT.

OK.

> In the Syntax document, in the definition of datatype maps, N_FA is a function mapping from
> datatypes to the set of that datatype's facet-literal pairs. I suppose that V_FA should therefore
> be a set of /datatype/-facet-literal /triples/.

Good catch -- this was an error. V_FA should, of course, also be a function; furthermore, it should state that there are no facets
for rdfs:Literal. I've changed the definition accordingly.

> I am unsure whether n-ary datatypes are members of the datatype map, too. If so, the above item
> must be changed to "for each datatype, except n-ary datatypes", since n-ary datatypes consist of
> /tuples/ of instances of the data domain (see semantics for ComplementOf(DR)!)

The way this comment is to be addressed depends on the outcome of the general discussion about n-ary datatypes. Here is, however,
what the correct interpretation should be in my opinion:

- Datatypes should not be n-ary. Calling a predicate such as < a datatype is bound to confuse people. The term "n-ary datatypes"
stems from the fact that you can treat all predicates with a built-in semantics in the same way from an implementation point of
view.

- Data ranges should be n-ary. If I understood the proposal for n-ary datatypes (for lack of a better word at the moment) correctly,
I believe that comparison and all other operations actually subclass the DataRange class from the MOF metamodel.

The way the semantics has been currently defined is coherent with this viewpoint. If we stick with this, we should probably say that
data ranges are to be interpreted as n-ary relations over \Delta_D.

For the moment, I'll add an editorial comment to the document.

> The semantics for ReflexiveProperty and IrreflexiveProperty are both given in the form
>
> FORALL x : x in DELTA_Int implies ...
>
> Nowhere else in the tables I can see the redundant expression "x in DELTA_Int" be used. I suggest
> to simply write
>
> FORALL x : <x,x> in (OPE)^OP

I agree with Markus here and would prefer to keep the table as is. In this way it is clear what set the variables range over.

> Maybe this and the last two sections should be reorganized. The section on "Ontologies" already
> tells what it means that an interpretation satisfies an ontology. The section on "Models" seems
> redundant to me (unless I missed something), since it just introduces a synonym: models are
> satisfying interpretations.

As Markus has noted, the present structure is deliberate. Let Int be an interpretation satisfying O. Clearly, Int is a model of O.
However, each interpretation Int' that differs from Int only in the interpretation of anonymous individuals is also a model of O.
I've added two sentences about this in Section 2.4.

> Typo: Is: "if a^I in C^CE". Must be: "CE^C".

Yes, sorry!

> Typo: The superprint of CE is CE, but should be C.

Yes, sorry!

> C^prime is not defined.

Yes, I messed it up. I've modified this part of the proof.

Received on Monday, 8 September 2008 09:13:40 UTC