Re: SKOS proposal - labels and notes

There is a saying that only God knows the true name of something. :-)

My experience with prefLabels is that they need to be assigned to a context in a mesh-up environment. Datasets that don't do this a priori need to be mapped to a context before being merged. In an RDF environment, this is clunky but possible by adding a skos:inScheme property when labeling concepts. 

If you want to assign prefLabels to non-conceptual things, I've found foaf:focus useful. That's what we do in VIAF. These extra levels of abstraction work, but they aren't easily understood.

Jeff

Sent from my iPad

> On Oct 20, 2013, at 6:35 PM, "Dan Brickley" <danbri@google.com> wrote:
> 
>> On 20 October 2013 17:30, Karen Coyle <kcoyle@kcoyle.net> wrote:
>> On 10/20/13 7:25 AM, Antoine Isaac wrote:
> 
>>> Would this imply that the recommendation would be to have only one
>>> schema:name, for the preferred label, and use altLabel and hiddenLabel
>>> for variants? It may be conflicting with the possibility (already used?)
>>> to have several schema:name for something.
>> 
>> 
>> Hi, Antoine. Thanks for the detailed explication of this issue.
>> 
>> As someone with a library background, I perhaps have a particularly strict
>> interpretation of SKOS. In my field, thesauri exist in equal part to create
>> a map of concepts and to control the terminology used for those concepts.
>> This latter function is less common in today's web environments, where
>> discovery is done on natural language, not controlled terms. Because there
>> are still KOS systems that use controlled terms as concept identifiers, I
>> would hate to see SKOS preflabel re-used in a less precise manner.
>> 
>> I believe this supports maintaining a distinction between schema:name and
>> skos:prefLabel, and not equating the two.
> 
> http://schema.org/name is currently defined as 'The name of the item.'
> (you might argue it should say 'a name' not 'the name', but setting
> that aside for now)
> 
> The way RDF and schema.org works, means that anything that is 'the
> name of the item' is a legit value for this property. Are there any
> values for a SKOS-like prefLabel that would not also count as 'a name
> of' the item/concept? To the extent concepts have names at all, I'd
> guess their preferred labels would all be names.
> 
> If not, i.e. if every preferred label of a concept is also a name, and
> if we still want to maintain an explicit notion of 'preferred label',
> then this seems a good candidate for describing as a sub-property /
> super-property relationship. We've used that notion already in the
> Action design, to relate focussed action-type-specific properties to
> the broader, vaguer properties on http://schema.org/Action. It might
> help here too (even though schema.org term navigation doesn't offer
> any support for sub-property links yet).
> 
> Dan
> 
> 

Received on Sunday, 20 October 2013 23:49:04 UTC