ISSUE-32: ConceptSchemeLabellingInteractions

ISSUE-32: ConceptSchemeLabellingInteractions

http://www.w3.org/2006/07/SWD/track/issues/32

Raised by: Alistair Miles
On product: SKOS

Can two different concepts in the same concept scheme have any lexical labels in
common?

I.e. can two different concepts in the same concept scheme both have the same
preferred lexical label? Can two different concepts in the same concept scheme
both have the same alternative lexical label? Can a lexical label be preferred
for one concept and alternative for a different concept in the same concept
scheme? Can a lexical label be hidden for one concept and either preferred or
alternative for a different concept in the same concept scheme?

See also:
<http://www.w3.org/2006/07/SWD/wiki/SkosDesign/ConceptSchemeLabellingInteractions>

Imagine the SKOS vocabulary consists of only skos:prefLabel, skos:altLabel,
skos:hiddenLabel, skos:Concept, skos:ConcepScheme and skos:inScheme.

Now consider the following graph:

@prefix ex: <http://www.example.com/examples#>.
@prefix skos: <http://www.w3.org/2004/02/skos/core#>.

ex:fooscheme rdf:type skos:ConceptScheme.

ex:A rdf:type skos:Concept.
ex:A skos:prefLabel "orange"@en.
ex:A skos:scopeNote "The colour orange."@en.
ex:A skos:inScheme ex:fooscheme.

ex:B rdf:type skos:Concept.
ex:B skos:prefLabel "orange"@en.
ex:B skos:scopeNote "The citrus fruit."@en.
ex:B skos:inScheme ex:fooscheme.

This graph is not inherently inconsistent.

However, if we were to assume that ex:fooscheme was a thesaurus, and attempt to
generate a thesaurus from this graph, following the conventions and notation of
ISO 2788 and BS 8723, we would have a problem. The problem is that two concepts
in the ex:fooscheme concept scheme have the same preferred lexical label, and we
can't tell from the graph whether these two concepts are supposed to be
different or not.

If we assumed that ex:A and ex:B were different concepts, we could not generate
a thesaurus, because it is an axiom of thesaurus construction that each
preferred term in a thesaurus has one and only one meaning, i.e. there is a
one-to-one correspondance between preferred terms and concepts.

If we assumed that ex:A and ex:B were the same concept, we would get the
following thesaurus:

orange
  SN The colour orange.
  SN The citrus fruit.

... which obviously violates the original intention.

Consider a different example:

@prefix ex: <http://www.example.com/examples#>.
@prefix skos: <http://www.w3.org/2004/02/skos/core#>.

ex:fooscheme rdf:type skos:ConceptScheme.

ex:A rdf:type skos:Concept.
ex:A skos:prefLabel "orange (colour)"@en.
ex:A skos:altLabel "orange"@en.
ex:A skos:scopeNote "The colour orange."@en.
ex:A skos:inScheme ex:fooscheme.

ex:B rdf:type skos:Concept.
ex:B skos:prefLabel "orange"@en.
ex:B skos:scopeNote "The citrus fruit."@en.
ex:B skos:inScheme ex:fooscheme.

Again, this graph is not inherently inconsistent.

However, problems arise when trying to generate a thesaurus from this graph. In
this case it is simply not possible to generate the thesaurus under any
assumption, because the lexical label "orange"@en would be mapped to both a
preferred term and a non-preferred term. It is an axiom of thesaurus
construction that a term must be either preferred or non-preferred.

Consider a third example:

@prefix ex: <http://www.example.com/examples#>.
@prefix skos: <http://www.w3.org/2004/02/skos/core#>.

ex:fooscheme rdf:type skos:ConceptScheme.

ex:A rdf:type skos:Concept.
ex:A skos:prefLabel "orange (colour)"@en.
ex:A skos:altLabel "orange"@en.
ex:A skos:scopeNote "The colour orange."@en.
ex:A skos:inScheme ex:fooscheme.

ex:B rdf:type skos:Concept.
ex:B skos:prefLabel "orange (fruit)"@en.
ex:B skos:altLabel "orange"@en.
ex:B skos:scopeNote "The citrus fruit."@en.
ex:B skos:inScheme ex:fooscheme.

Again, this graph is not inherently inconsistent.

Again, however, problems arise when trying to generate a thesaurus from this
graph, which are summed up in SkosDesign/ThesaurusPatterns.

Summary

Thesauri are a major application of the SKOS vocabulary. However, to be usable
as a thesaurus, an RDF representation has to observe some syntactic constraints
on the usage of the SKOS vocabulary. These syntactic constraints should probably
be optional, because other types of vocabulary may not need them, but we still
have to say what they are.

Received on Monday, 19 March 2007 14:23:54 UTC