[open-bibliography] New BNB sample data available

Ford, Kevin kefo at loc.gov
Fri Feb 4 22:45:16 UTC 2011


"One thing to be careful of with prefLabel is that SKOS does not explicitly express the constraint that  prefLabel's must be unique in a given scheme."

Small point, but I'm not certain this is implied either.  In other words, I don't think there is comment one way or another about two distinct resources (concepts) with identical prefLabels in the same scheme.  The "pairwise disjoint" rule applies to one concept (a resource) having no more than one prefLabel per language.  (Actually...because a skos:Concept can belong to zero or more skos:ConceptSchemes, one might argue that the designers accepted the notion that two distinct concepts that share an identical prefLabel might one day end up in the same scheme.  Implying, of course, that two distinct concepts with identical prefLabels in the same scheme was an acceptable possibility.  There's still nothing that explicitly or implicitly suggests they are owl:sameAs to each other.  But, I digress.)

As you know, in SKOS, when representing LC subject headings, which come from the MARC, we lose the distinctions between the components of a subject heading - the fact that one component is a Topic, another a Geographic Subdivision, something a Genre/Form, etc.  In MADS/RDF, the componentList components, which represent the individual headings that constitute the authoritativeLabel, will indicate the types of the individual headings found in the authoritativeLabel.  So, there might be two resources in the same scheme with identical authoritativeLabels, but the resources collected in the MADS/RDF componentList provide a means to further disambiguate between the two headings.  But, like SKOS, there is no attempt to say that these data points for any one resource must be unique in a given scheme. ( The fact that a particular heading is a subdivision is information that will be recorded differently.  MADS/RDF does not attempt to capture such information explicitly. )

Cordially,

Kevin




________________________________________
From: public-lld-request at w3.org [public-lld-request at w3.org] On Behalf Of Simon Spero [ses at unc.edu]
Sent: Friday, February 04, 2011 16:42
To: Antoine Isaac
Cc: Young,Jeff (OR); open-bibliography at lists.okfn.org; public-lld
Subject: Re: New BNB sample data available

On Thu, Feb 3, 2011 at 4:46 PM, Antoine Isaac <aisaac at few.vu.nl<mailto:aisaac at few.vu.nl>> wrote:
Hi Jeff,

As this coordination issue is complex indeed, this was not handled in SKOS. Instead it relies on specific extensions to be designed--this is MADS/RDF [1] is made for, in fact!

There are complexities in the syntax and semantics of subdivided, coordinated, compound, and faceted headings that aren't really addressed by the MADS/RDF proposal.

I have a presentation covering  some  aspects of the semantics of subdivided and faceted headings at ISKO-UK conference  in London this year.

One thing to be careful of with prefLabel is that SKOS does not explicitly express the constraint that  prefLabel's must be unique in a given scheme.  This restriction is tricky to express, even with OWL 2  Key Properties, as prefLabels are annotation properties, and are thus not covered by the OWL 2  semantics.  The constraint can be expressed for skosxl, as skosxl:prefLabel is an ObjectProperty, and literalForm is a DataProperty (though there's a little bit of trickiness needed to make sure that two Labels with the same literalForm aren't used as skosxl:prefLabel for different concepts).  Also, if the same label is used for two concepts, HasKey will infer that they are owl:sameAs.

Simon




More information about the open-bibliography mailing list