On 4/11/12 3:50 AM, Thomas Baker wrote:
> On Tue, Apr 10, 2012 at 08:39:28PM +0200, Antoine Isaac wrote:
>>> Can we say that ontologies constrain a conceptual universe ("Things"), and
>>> application profiles constrain a data representation ("Data")? The constraints
>>> in SKOS were about the SKOS "model", not about data created in accordance with
>>> that model.
>> I don't buy that :-)
>> SKOS is a data model, we did not create these constraints not looking at the
>> possible data. As a matter of fact the constraints are examplified by data
> Okay, then, let me rephrase that...:-) Consider the following typical example,
> illustrating the constraint that a concept has only one prefLabel per language
> The following graph is not consistent with the SKOS data model, because two
> different preferred lexical labels have been given with the same language
> Example 12 (not consistent)
> <Love> skos:prefLabel "love"@en ; skos:prefLabel "adoration"@en .
>  http://www.w3.org/TR/skos-reference/
> To me, this only says that these two RDF statements, taken together, would not
> be consistent with the SKOS data model. Yes, one would want to understand this
> in order to design a particular concept scheme well, and one could use this to
> "validate" a particular concept scheme. However, under the OWA, those triples
> could have come from two different sources, and they would be just as
> inconsistent with the SKOS data model.
Yes. This depends on what one defines as "one's data" to validate.
> I see these constraints as being
> defined under open-world assumptions, though I guess they could easily used as
> closed-world constraints... (For those who know Pellett: Does this fit with
> the Pellett idea?)
Yes in general, no in particular. That one axiom is indeed not really dependent on OWA, I think: as soon as you've got two prefLabels with a same language tag, the graph is inconsistent. Considering that you may have data somewhere else, or in the future, wouldn't change anything.
>>> Agreed. However, it is perhaps useful to contrast "open-" and "closed-world"
>> It is, yes!
> Is it black-and-white, or are there shades of grey?
The latter, as always ;-) And I'd be reluctant to discuss the shades of grey here, right as the RDF group is discussing Named Graphs. Let's just gather the DCMI cases first, and fire them at whatever draft or spec the RDF group has then!