RE: URI thoughts

> Assuming I have a RDF model of the follows.
>
> @prefix rdf: http://www.w3.org/1999/02/22-rdf-syntax-ns#.
> @prefix dctype: http://dublincore.org/2003/12/08/dctype#
>
> <http://www.charlestoncore.org/images/group-logo.png> rdf:type dctype:Image.
>
> A simple model without any closure will contain only this one statement.  It
> is not very useful, because given this model, I cannot even write a program
> to tell a human that the resource
> http://www.charlestoncore.org/images/group-logo.png is a "type" of "Image".
> Note, the type and Image are annonated in their respective RDF document via
> rdfs:label.  Without dereferencing the URI, you won't be able to obtain how
> to label the graph and discribes it to a human.
>
> Different types of closure gives you different scope of knowledge base, from
> which different answers may be given to the same query.

Absolutely, what I'm suggesting is a framework where the extent / scope of 
the closure is designated (to some extent) by the author of the instance 
graph as well as the authors of defining ontologies.  In which case, the RDF model above 
would instead be:

@prefix rdf: http://www.w3.org/1999/02/22-rdf-syntax-ns#.
@prefix dctype: http://dublincore.org/2003/12/08/dctype#.
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>.

<> rdfs:isDefinedBy <http://purl.org/dc/dcmitype/>.

<http://www.charlestoncore.org/images/group-logo.png> rdf:type dctype:Image.

Note the difference in the dctype base uri in the instance document and 
the URL of the defining RDF schema.  <http://purl.org/dc/dcmitype/> is a 
PURL.org URL which redirects (currently) to 
<http://dublincore.org/2003/12/08/dctype#>.

The former is the 'authoritative' location of the defining ontology which wouldn't change 
but could redirect to an RDF graph of the owner's choice (with different 
URIs for - perhaps - a later version of the same concept).  Relationships 
between versions of the same concept can be maintained with appropriate 
assertions in the defining ontologies using dc:replaces and dc:isReplacedBy 
(as pointed out earlier).

Following this convention recursively, the The DCMI Types schema 
*could* (it currently doesn't) provide links to the schemas for 
all the primary vocabulary terms it uses (RDF,RDFS,DC,DCTerms) with 
similar rdfs:isDefinedBy statements.  Eventually this somewhat controlled 
recursion ends up with OntologicalClosure without a brute force attempt to 
dereference *all* class and property URIs.

Interestingly, in browsing the DCMI Type RDFS schema I noticed other 
vocabulary terms that 'suggest' dereferencing to RDF:

- dcterms:requires
- dcterms:isRequiredBy

But as with rdfs:seeAlso, they are sometimes used to link resources for 
human consumption (HTML documents) and other times used to link RDF graph 
URLs.

I think I'll make an attempt to flesh this out on the ESW wiki.

> I have tried to explain in the past the importance of ontology modulization.
> Building a monolithic ontology, i.e., by putting all concepts under one
> namespace, is VERY dangerous because dereferening one URI will in fact
> retrieve all RDF statements under the same namespace. This will make the
> sharing of ontology difficult.
>
> This is also what prompt me to propose the concept of ontology
> normalization.  But to avoid self-promotion, I will leave it to private
> discussion if anyone is interested.
>
> Xiaoshu
>
>
Chimezie Ogbuji
Lead Systems Analyst
Thoracic and Cardiovascular Surgery
Cleveland Clinic Foundation
9500 Euclid Avenue/ W26
Cleveland, Ohio 44195
Office: (216)444-8593
ogbujic@ccf.org

Received on Tuesday, 20 June 2006 20:30:40 UTC