Karen Coyle wrote:
> On 3/14/12 2:36 PM, Tom Baker wrote:
>
> >
> > Is it clearly "best practice" to use one of these variants? Or are
> there
> > circumstances in which one might favor one over the other?
>
> I think it depends entirely on what you *have* rather than what you
> *should have*. So perhaps it would be good to state this as: if you
> have
> 'x' do 'y'.
+1
>
> I am unclear on the use of skos:prefLabel in instance data. I consider
> skos:prefLabel to be appropriate for the definition of vocabularies. I
> can imagine a system carrying labels locally for reasons of
efficiency,
> but I would not want to see instance data over-riding the preferred
> label defined in a vocabulary... or is this a feature, not a bug?
In effect, skos:prefLabel is the W3C's admission that libraries invented
"Linked Data" first. :-)
The problem is that we didn't come up with the idea of URL-encoding
"controlled access points" and prefixing them with the letters
"http://". We (actually human-beings in general) also have this bad
habit of changing our minds about the persistenc of
prefLabels/identifiers (aka "names"). Thus LCCNs...
skos:Concept/skos:prefLabel/skos:inScheme are important in instance data
because they allow others to connect their legacy data to an emerging
(and profoundly philosophical) consensus of reality.
>
> >
> > Would it be useful to move these examples to a higher level of
> abstraction --
> > i.e., express these examples not only as RDF triples, but in XML
> Schema or
> > Schematron?
>
> I see no reason not to show different serializations of the examples,
> but it may to be stretch to intend to show them all in all
> serializations.
I agree.
>
> >
> > If the potential value of DCAM lies in providing a frame for
examples
> of Best
> > Practice, is it more important to get the set of examples right or
to
> get the
> > more generalized abstract syntax right?
>
> If DCAM is re-engineered as a set of best practices then I'm not sure
> that there *is* an "abstract syntax" per se. The abstract syntax may
be
> RDF+OWL/RDF+SKOS, etc.
RDF/OWL/SKOS represent different levels of abstraction. This is a crude
approximation, but RDF is a vocabulary that allows us to create
"statements".
OWL/RDFS is a domain-neutral vocabulary that allows us to define
domain-specific vocabularies.
SKOS is a "simple" domain-specific (context-sensitive) "prefLabel"
vocabulary that some situations require for people/systems to "get their
bearings". See foaf:focus/madsrdf:isIdentifiedByAuthority for
standardized relationships to reality.
FOAF, Schema.org, DC Terms, BIBO, RDA, and many others are attempts to
create domain-specific (generally "reality-based") vocabularies.
>
> In the case above, there is more than
> > one way to do it. Is it perhaps enough to assemble useful examples,
> starting
> > with simple patterns like those above, on up to higher-level
patterns
> like the
> > one Dan presents?
>
> I think it's an interesting exercise to postulate a set of simple
> examples as well as higher-level patterns, in part to determine if the
> higher-level patterns require more than a compound of the simple
> patterns. I definitely think that assembling patterns is a useful
> exercise. An interesting example of complex patterns could be MADS,
> which attempts to re-create in RDF the full MARC authority record.
MADS maps to SKOS (see above).
>
> http://www.loc.gov/standards/mads/rdf/
>
> This is a real-life example with plenty of complexity.
Disclaimer: I'm not claiming that anyone thinks like this except me.
Jeff
>
> kc
>
> >
> > Discuss...
> >
> > Tom
> >
>
> --
> Karen Coyle
> [log in to unmask] http://kcoyle.net
> ph: 1-510-540-7596
> m: 1-510-435-8234
> skype: kcoylenet
|