Personally, I'm not a fan of the relators being defined as both skos:Concept and as rdf:Property. I believe that this was done to accommodate earlier use of MARC Relators as refinements of dc:contributor [1], and the desire to express them as skos:Concept. I'm not exactly sure what the advantage of having them as skos:Concept. So I'd like to see some examples of how it is supposed to be used as well. I'm also not directly involved in the project anymore, so I imagine someone from the Network Development and MARC Standards Office might have more information.
//Ed
[1] http://dublincore.org/usage/documents/relators/
________________________________________
From: Authorities and Vocabularies Service Discussion List [[log in to unmask]] On Behalf Of Bernard Vatant [[log in to unmask]]
Sent: Thursday, May 27, 2010 6:32 AM
To: Summers, Ed; [log in to unmask]
Subject: skos:Concept AND rdf:Property : what is the rationale?
Hi all
I've suscribed to this list following a today exchange on Twitter between
Gautier Poupeau and Ed Summers regarding the new vocabularies published at
id.loc.gov, and in particular at http://id.loc.gov/vocabulary/relators,
where all "relators" are defined both as instances of skos:Concept and
rdf:Property.
I understand the idea behind this.
Any relator (aka "role") is by nature ambiguous in natural language.
"Composer" can be viewed as a class ...
<http://dbpedia.org/page/Wolfgang_Amadeus_Mozart> rdf:type "composer"
(Mozart is a Composer)
.. or as a property ...
http://dbpedia.org/page/Don_Giovanni "composer"
<http://dbpedia.org/page/Wolfgang_Amadeus_Mozart> ...
or as a Concept
<http://opera.stanford.edu/composers/> dc:subject "composers".
It is considered good practice for a semantic web vocabulary when it comes
to representing such ambiguous concepts, to make a choice, based on the
functional target of the vocabulary, in order to avoid driving logical tools
(and human ontologists) crazy, and to control the functional use of the
concept (classifying things, declaring relationships, or indexing resources).
But another (supportable IMO) viewpoint could be to declare ambiguity in the
vocabulary, and let the applications sort the ambiguity out. Seems that this
is the option taken by the relators vocabulary.
Since people have frowned at this, could soemeone (Ed, I guess) explain the
rationale, and maybe document examples of use of a relator as a skos:Concept
and as a rdf:Property.
Bernard
|