Re: Interpretation of RDF reification

From: Henry Story <henry.story@bblfish.net>
Subject: Re: Interpretation of RDF reification
Date: Thu, 23 Mar 2006 16:57:44 +0100

> 
> On 23 Mar 2006, at 16:44, Peter F. Patel-Schneider wrote:

[Context added back in ...

From: Henry Story <henry.story@bblfish.net>
Subject: Re: Interpretation of RDF reification
Date: Thu, 23 Mar 2006 16:36:35 +0100

> On 23 Mar 2006, at 15:58, Peter F. Patel-Schneider wrote:
> > Date: Thu, 23 Mar 2006 13:47:17 -0000
> >>>
> >>> I don't understand the difference.  RDF just does not have the  
> >>> expressive
> >>> power to do this sort of thing.   It just like asking whether  
> >>> propositional
> >>> logic could express something like "All students are people."
> >>
> >> I think that answers my question.
> >> I'm confused about how semantics of [OWL] relate to the semantics  
> >> of RDF.
> >
> > Roughly in the way that propositional logic relates to predicate  
> > logic (at
> > least if you think of propositional logic in a certain way).  There  
> > are
> > "more" constructs in OWL, and these constructs have meaning that  
> > cannot be
> > expressed in RDF.  The only strangeness is that the extra OWL  
> > constructs
> > are actually written as (collections of) RDF triples, but  
> > nonetheless the
> > added expressive power is still real.

...]

> >> I hope that is wrong.
> >
> > Nope, it is correct.
> 
> What do you[r] base you position on?

OWL Web Ontology Language
Semantics and Abstract Syntax
W3C Recommendation 10 February 2004 
http://www.w3.org/TR/owl-semantics/

which provides a detailed account of the meaning of the OWL constructs.
Actually there are two accounts there - the one most germane to this
dicussion is in Section 5. RDF-Compatible Model-Theoretic Semantics
(Normative), available as a separate document at

http://www.w3.org/TR/owl-semantics/rdfs.html

> >> My understanding is rather that RDF gives you a
> >> framework.
> >
> > Well, just *what* is a framework?
> 
> It states the basic elements and how they can be put together.

Well, then I guess everything is a framework.

> >> OWL just gives you specialised relations with certain specific
> >> inferential properties.
> >
> > Not really.  OWL in RDF gives you much more than specialized  
> > relations.
> > Much of the power of OWL comes from syntax that is more than just  
> > single
> > relations.  It so happens that it is possible to embed OWL in RDF in a
> > certain manner.  (This is not always possible, by the way, and OWL  
> > is very
> > close to the maximum expressiveness that can be so embedded in RDF.)
> 
> I don't think anything in OWL depends on the syntax. I can write own  
> triples in rdf/xml,
> N3, Turtle, etc...

Well, of course, what matter in RDF (and OWL) is just the triples, and none
of the above depends on the surface syntax of the triples - so long as you
talk only about triples.

> >> So if you take the relation owl:inverseOf then this is just an [RDF]
> >> relation.
> >
> > Well, sure, owl:inverseOf is *just* an RDF property, in RDF.  In  
> > OWL, on
> > the other hand, owl:inverseOf is a special property - it has a extra
> > meaning provided by the OWL semantics.
> 
> This is where it would be interesting to understand if this  
> difference in semantics is
> complementary or widely different. My guess is that it is  
> complimentary. owl relations
> have special semantics, just as the word "brother" in english has a  
> specialised meaning without nevertheless ceasing to be a word. Now in  
> rdf we can create a fam:brother relationship, and it will just happen  
> that if you look at the universe of discourse this fam:brother  
> relationship is indeed what we know in english as being transitive.  
> owl just
> gives us a word for that: owl:transitive

Well, the OWL Full semantics is indeed an extension of the RDF semantis,
so any OWL Full meaning given to an RDF graph is an extension of the RDF
meaning given to that same graph.

> >> But it is linked to the following well known rule:
> >>
> >> { ?r1 owl:inverseOf ?r2 .
> >>    ?a ?r1 ?b .  } => { ?b ?r2 ?a . } .
> >
> > Not all all.  There are *no* rules in RDF, nor in OWL.  The above  
> > is not
> > even legal RDF syntax, nor legal OWL syntax.
> 
> There may be no rules in [RDF], but that does not stop it from being  
> possible to create other
> languages that have rules. Rules just are short hands for stating  
> patterns of relationships. And these patterns exist whether they can  
> be expressed in  a well know way in the language or not.

Well, perhaps, but you do then need to do much more to say just what the
syntax above means.  In fact, there is a document precisely says what the
extra OWL meaning of owl:inverseOf is.  Not very surprisingly the document
is:

OWL Web Ontology Language
Semantics and Abstract Syntax
W3C Recommendation 10 February 2004 
http://www.w3.org/TR/owl-semantics/

The relevant part is taken from Section 5.2. OWL Interpretations in the
table titled "Characteristics of OWL vocabulary related to equivalence"

If E is 	then <x,y> in EXTI(SI(E)) iff

owl:inverseOf  	x,y in IOOP and <u,v> in  EXTI(x) iff <v,u> in EXTI(y)

This may, or may not, have the same effect as some reading of the bit of
syntax that you provide above.

> >  All of the [OWL] vocabulary can be defined in terms of such rules.  
> > The
> >> OWL terms have been very carefully selected to make certain types of
> >> inferencing easier. But otherwise OWL just is a specialised  
> >> vocabulary.
> >
> > Not at all.  OWL is a a different logical language from RDF.  It just
> > happens that OWL (Full) was designed as a same-syntax extension of RDF
> > Schema.
> 
> I don't believe that. Where is that written?

OWL Web Ontology Language
Semantics and Abstract Syntax
W3C Recommendation 10 February 2004 
http://www.w3.org/TR/owl-semantics/

provides a full semantic account of OWL, separate from the semantic account
of RDF, which is in

RDF Semantics
W3C Recommendation 10 February 2004
http://www.w3.org/TR/rdf-mt/


peter

Received on Thursday, 23 March 2006 16:22:09 UTC