W3C home > Mailing lists > Public > public-swd-wg@w3.org > March 2008

[SKOS] Story of label relations

From: Thomas Baker <baker@sub.uni-goettingen.de>
Date: Mon, 24 Mar 2008 10:20:33 +0100
To: SWD Working Group <public-swd-wg@w3.org>
Message-ID: <20080324092033.GA3696@octavius>

Hi all,

I'm not quite understanding the story we tell about the
n-ary relation solution to label relations.

The Primer says that relationships between labels are treated
as first-order RDF resources.  Each instance of this class,
with its associated literals, may then ("later") be connected
to concepts using skos:seeLabelRelation.

About seeLabelRelation, SKOS Reference says simply:
    
    skos:seeLabelRelation rdfs:domain rdfs:resource . [1]

and about the labels which are related, SR says only:

    "The rdfs:range of skos:labelRelated is the class of RDF plain literals." [3]

So if we take the graph:

    ex:FAO skos:seeLabelRelation [ skos:labelRelated "baz"@en , "quux"@en ] .

we have no basis to infer:

    ex:FAO rdf:type skos:Concept.

or even:

    ex:FAO rdfs:label "baz"@en ;
           rdfs:label "quux"@en .

Indeed, 8.6.2. [2] emphasizes that "there does not necessarily
have to be any correspondence whatsoever between the lexical
labels of a resource, and the labels involved in an associated
label relation".  Therefore, the following is consistent:

    <FooResource>
      skos:prefLabel "foo"@en ;
      skos:altLabel "bar"@en ;
      skos:seeLabelRelation [ skos:labelRelated "baz"@en , "quux"@en ] .

If the association of an instance of the label relation class
with particular concept is arbitrary, I'm also seeing no reason
_not_ to take a label relation connected to <QuuxResource>
as a statement about relations between labels of <FooResource>:

    <FooResource>
      skos:prefLabel "foo"@en ;
      skos:altLabel "bar"@en .
    
    <QuuxResource> 
      skos:seeLabelRelation [ skos:labelRelated "foo"@en , "bar"@en ] .

And I'm seeing no obvious reason why a label relation would not
equally apply to a concept in another concept scheme, as in:

    <FooResource>
      skos:prefLabel "foo"@en ;
      skos:altLabel "bar"@en ;
      skos:inScheme <MyScheme> .
    
    <QuuxResource> 
      skos:seeLabelRelation [ skos:labelRelated "foo"@en , "bar"@en ] ;
      skos:inScheme <AnotherScheme> .

(In Amsterdam, Antoine asked "what if we mix two concept
schemes? which scheme should we attach the label relationships
to?" [6]).

In sum, the Primer says that "some applications require the
creation of explicit links between the labels associated
to concepts".  

However, I do not understand whether the literals
that are explicitly linked in an instance of the label
relations class really are also explicitly being associated
with the labels of the concept that is the subject of the
seeLabelRelation statement.  

Or, alternatively, whether the instances of the label
relations class merely amount to global assertions about the
relationship between two literals in the world, wherever
character-for-character identical literals may be found -
whether linked to the same or different concept, within the
same or different concept scheme, or even in an entirely
non-SKOS context.  Are the label relation resources perhaps
intended as free-floating annotations that can be attached
here or there in a pragmatic way, as for display or documentary
purposes?

Moreover, Margherita's point still stands that the "n-ary
pattern" solution involves the duplication of information
because label literals are written twice [5], which is not
very pretty.

Tom
  
[1] http://www.w3.org/TR/2008/WD-skos-reference-20080125/#seeLabelRelation
[2] http://www.w3.org/TR/2008/WD-skos-reference-20080125/#L3235
[3] http://www.w3.org/TR/2008/WD-skos-reference-20080125/#L3059
[4] http://lists.w3.org/Archives/Public/public-swd-wg/2007Jun/0120.html
[5] http://www.w3.org/2006/07/SWD/wiki/SkosDesign/SKOS-XL?action=recall&rev=2
[6] http://www.w3.org/2007/10/09-swd-irc#T10-20-21

-- 
Tom Baker - tbaker@tbaker.de - baker@sub.uni-goettingen.de
Received on Monday, 24 March 2008 09:21:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 24 March 2008 09:21:09 GMT