W3C home > Mailing lists > Public > semantic-web@w3.org > September 2010

RE: Reuse

From: Armando Stellato <stellato@info.uniroma2.it>
Date: Tue, 21 Sep 2010 22:54:18 +0200
To: "'Peter Williams'" <pezra@barelyenough.org>, <semantic-web@w3.org>
Message-ID: <06e401cb59cf$2bd9d6b0$838d8410$@uniroma2.it>
Hi Peter,

Just my poor cent on this...

There have been probably historical (and sometimes practical) reasons why
these overlaps happened to be:

A few of the ontologies you listened were originated from porting already
existing standards (e.g. Dublin Core, vcard etc..) in RDF, and were intended
to provide a minimal but still complete coverage of these standards, thus
without importing whole other ontologies.
In other cases, to create connections with other commonly used knowledge
bases, links were created (through rdfs:subClassOf rels) without explicitly
importing the ontology, such as in previous FOAF versions, where references
to the first RDF porting of the WordNet lexical database were included. The
reason behind the links and the repetitions of the concepts (such as
foaf:Person rdfs:subClassOf wn:Person) were probably dictated by practical
reasons (if any person using FOAF had to import all the hundreds of Mbytes
of RDFWordnet...).

Also, there have been proposals for URI repositories, helping to avoid
proliferation of URIs in the Web, such as: 
http://www.okkam.org/
though in that case they mostly refer to objects rather than concepts.

Linked Data Initiative provides some guidelines on this:
http://www4.wiwiss.fu-berlin.de/bizer/pub/LinkedDataTutorial/#whichvocabs

So, in the end, well, my simple advice is: those all-star ontologies may
contain some overlap, but they reside in the Olympus of ontologies (and they
are in any case interconnected, so, whichever you link, you get your data
"open-linked") and they can :-).
Whenever you need to represent concepts in your ontology, reuse as much as
possible concepts from those existing ones. You may make exceptions just on
practical basis, i.e. if you need one specific concept from a OO (Olympus
ontology :-) ) and only that one, and that OO contains lot of other
information you really don't need (that is, you don't want to import it),
then define your own concept in your ontology, and then put a triple with
equivalentClass or subClassOf between your concept and the "famous one".
Also apply the same consideration to properties and, most of all, the
combination of both (that is, in the example above, instances of that single
concept should not be described in your ontology properties contained in the
OO for your practical consideration to be valid).

Cheers,

Armando






> -----Original Message-----
> From: semantic-web-request@w3.org [mailto:semantic-web-request@w3.org]
> On Behalf Of Peter Williams
> Sent: Tuesday, September 21, 2010 8:02 PM
> To: semantic-web@w3.org
> Subject: Reuse
> 
> I have a question regarding reuse in rdf vocabularies.
> 
> There seem to quite a few examples of vocabularies that reproduce
> concepts that exist in other vocabularies.  For example,
> 
>  * dc:Agent and foaf:Agent
>  * dc:LicenseDocument, cc:License, doap:License and probably half a
> dozen other license classes
>  * every thing in <http://www.w3.org/Submission/vcard-rdf/> and foaf.
> 
> I am sure i could keep going but you get my point.
> 
> What is considered best practice regarding reuse when designing new
> vocabularies?  Should existing vocabularies be used?  If so under what
> circumstances?  In which situations is defining classes and properties
> that overlap with existing vocabularies advantageous?
> 
> When faced with multiple vocabularies describing the same thing how do
> you choose which to reuse?
> 
> Peter
Received on Tuesday, 21 September 2010 20:54:48 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:38 GMT