Re: Question about value vocabularies and range

Hi Karen

I get that you may be looking for a wider range of opinions, but until they
weigh in...

 :-)

On Fri, May 6, 2011 at 11:43 AM, Karen Coyle <kcoyle@kcoyle.net> wrote:

>  Thanks to Jeff, Kevin and Jon.
>
>
> > Basically the owl restriction
> > owl:allValuesFrom says that any value for "Language of text" must
> considered
> > to be of a particular type (member of a specific class).
>
> I probably wasn't specific enough in my question, though.
> dc:LinguisticSystem, refers to any linguistic system, not a specific
> list, right?
>

If you think of dc:LinguisticSystem as a SuperClass, then any owl:Class that
subclasses dc:LinguisticSystem, as Lingvoc does, would be consistent with a
property that has an owl:Restriction
of owl:allValuesFrom dc:LinguisticSystem.

We can't ignore the effect of inferencing on 'range' restrictions.


> I think a class definition is too broad for what I'm
> needing. (Which is why I suspect this may get into the application
> profile area.)
>

So, I'm a bit confused by the distinction you're making between APs and OWL,
since APs are distinctly human documentation and OWL is (at least
potentially) a machine-readable expression of an AP.

Also, there aren't any realistic restrictions on the nature, size, or
membership of an owl:Class in this context. There's no reason why <
http://id.loc.gov/vocabulary/iso639-2> wouldn't be considered by an
owl:AllValuesFrom restriction to be a distinct owl:Class, even if it wasn't
explicitly declared to be any other type than a skos:ConceptScheme (itself
an owl:Class). You've simply implied that any values for that property must
be considered to be members of the owl:ConceptScheme identified by <
http://id.loc.gov/vocabulary/iso639-2>.


>
> owl:allValuesFrom can either define a class, or a datatype. To specify
> a list of values ("red, blue, green") you can use owl:oneOf, and the
> values then follow... but can you say: "owl:oneOf ->
> http://RDVocab.info/termList/RDAcolour?"
>
>
After reading <http://www.w3.org/TR/owl-ref/#EnumeratedDatatype> I don't
think so, at least not the way it's currently exported from the Registry as
a skos:ConceptScheme. But why would you want to?


> Maybe it would be best to take OWL out of the equation and ask:
>
> I want to define a property that takes a specific list of values as
> its range. That list of values needs to be defined independently of
> the property definition because it will be used by other properties as
> well and will be maintained separately from the property definition.
>
> What's the best way to do this?
>

It matters whether the 'values' are literals or non-literals, but I would
say generally as a distinct class that's a subclass of owl:Class.

And the more philosophical question:
> Should this be "relegated" to application profiles rather than
> building this constraint into a property definition?
>

And if you 'relegate' this to an AP, how do propose to express the
restriction for machines in an RDF context if not OWL or RDFS?

Jon


>
> kc
>
>
> --
> Karen Coyle
> kcoyle@kcoyle.net http://kcoyle.net
> ph: 1-510-540-7596
> m: 1-510-435-8234
> skype: kcoylenet
>
>