Re: Annotations and entailments

From: Jeremy Carroll <jjc@hpl.hp.com>
Subject: RE: Annotations and entailments
Date: Sun, 19 Jan 2003 21:18:19 +0100

> Peter Crowther:
> > Is it 'incorrect', though?  Seems to me this is the same problem as
> > whether the following two files are 'identical':
> 
> > FileA:
> > int main () { return 0; /* Do nothing */ }
> 
> > FileB:
> > int main () { return 0; /* Successful termination */ }
> 
> No, not at all.
> Classes in both RDF and OWL are intensional.
> 
> So two incomplete ontologies:
> 
> FileA:
> <owl:Class rdf:about="http://example.org/#People">
>    <rdfs:comment>Network Inference employees</rdfs:comment>
> </owl:Class>
> 
> FileB:
> <owl:Class rdf:about="http://example.org/#People">
>    <rdfs:comment>Expert ontologists</rdfs:comment>
> </owl:Class>
> 
> have different intensions before we start worrying about the extensions.

Hmm.  Ontologies don't have intensions, per se.  In fact, the only
difference betweeen these two ontologies is in the rdfs:comment.  If this
was removed, the two ontologies would be equivalent.

> While currently any OWL DL interpretation of the one is an interpretation of 
> the other, this is problematic precisely because of the failure to connect 
> with the real world; where the coincidence of "Expert ontologists" and 
> "Network Inference employees" is at best a fortuitous fact about this world 
> rather than a necessary truth about all possible worlds.

Argh.  This is buying into the social meaning doctrine of RDF with a
vengence.  As far as the RDF model theory is concerned, the only difference
is in the rdfs:comment tags.  Everything else about the meaning of the two
ontologies is the same.  In particular, in every interpretation, the nodes
in the two RDF graphs that is the class refer to the same resource, and
have the same class extension.  As far as being intensional sets goes,
then, the two nodes indeed have the same intension.

> The intent of the following in FIleC is unclear unless we import either FileA 
> or FileB (or both).
> 
> <eg:People>
>    <eg:name>Peter Crowther</eg:name>
> </eg:People>

Well, not in RDF (or OWL Full).  The above file has a perfectly clear
meaning and intent.  In the absence of a wholesale theory of social
meaning, which in the above case includes a theory of natural language
syntax, semantics, and pragmatics, adding FileA or FileB adds extrememly
little. 

> Hmm, maybe the C analogy isn't wholly mistaken - FileA seems to be finished, 
> whereas FileB appears to be a stub. 

Huh?  How can FileA be more ``finished'' than FileB?

> Calling the resulting program within a 
> shell script results in two very different situations. FileA may result in a 
> shell script that is either correct or not; whereas FileB would result in a 
> shell script that is not yet finished, because one if its parts still needs 
> work.

What part of FileB is unfinished?  It sure looks like FileB contains a
perfectly acceptable class to me.

> Jeremy

Peter F. Patel-Schneider
Bell Labs Research
Lucent Technologies

Received on Sunday, 19 January 2003 16:20:27 UTC