Re: TGN place types (broader/narrower spanning ConceptSchemes)

Hi Jutta!

14.04.2014 23:07, Jutta Lindenthal kirjoitti:

> I definitely would, but fortunately the GND is right:
> http://d-nb.info/050559028/about/rdf  [SG Dynamo Dresden]
[..]
>
> Well, the actual example from the GND exemplifies that the BTI relationhip
> in thesauri is envisaged to link individuals to their resp. generic classes
> (ISO 25964-1: "10.2.4.1 The instance relationship links a general concept,
>
> such as a class of things or events, and an individual instance of that class,
> which is often represented by a proper name [...]"; e.g.
> http://d-nb.info/standards/elementset/gnd#broaderTermInstantial).
>
>
> Individual instances in this context are usually named entities, also called
> classes-of-one, corresponding to FRBRoo "manifestation singleton", and may also
> include FRBRoo "manifestation product type". The ISO definition of the
> instance relationship appears slightly different from rdf:type, because the
> former is a semantic definition and the latter a formal one.

Thanks for the clarification. So it appears that the example was not 
correct and indeed the ISO standard has a stricter definition of where 
BTI can be used than what I thought.

It's a good thing to highlight how BTI still differs from rdf:type. Use 
of the latter in a thesaurus context would in effect be punning the 
"class concept" (the concept the BTI relationship points to) to be both 
a skos:Concept instance and an RDFS/OWL Class at the same time. That is 
a tricky terrain to navigate so defining a separate broaderInstantial 
relationship that avoids the baggage of rdf:type probably makes sense.

>>    by Osma Suominen (Aalto University), 11 May 2011
>>    has similar things: broaderGenericTransitive, broaderPartitiveTransitive.
>
> I wondered why these properties were explicitly defined, because generic hierarchies
> are transitive by definition. Most authors claim partitive relationships to be transitive
> as well, but transitivity between meronyms and holonyms only holds under certain conditions
> (the parts being unique, constituent and necessary parts of the whole).

I can't recall the specifics of what I thought back then, but even if I 
today had to define a new broaderPartitive (or broaderGeneric) property, 
which would be a subproperty of skos:broader, I would not make it an 
owl:TransitiveProperty. The reason is that it would easily create 
improper skos:broader inferences. Consider the following:


# property definitions
:broaderPartitive a owl:ObjectProperty, owl:TransitiveProperty ;
   rdfs:subPropertyOf skos:broader .

# asserted instance data
:fin a skos:Concept ;
   :broaderPartitive :europe .

:europe a skos:Concept ;
   :broaderPartitive :world .

:world a skos:Concept .


Now an OWL inference engine would conclude:

:fin :broaderPartitive :world .

...and further:

:fin skos:broader :world .


In effect, this would make skos:broader the same as 
skos:broaderTransitive for these cases so it would no longer mean "is 
immediately broader" (as in the excellent explanation by David Booth 
earlier in this list). Instead, there would be skos:broader 
relationships across all the hierarchy levels, making it rather hard to 
e.g. render the hierarchy in a usable way.

The solution, as in SKOS Core itself and in the FinnONTO SKOS extensions 
discussed above, is to make a transitive superproperty 
(broaderPartitiveTransitive) that can be used if we need to consider the 
full transitive closure. (Though I would not do that today, because it's 
an optimization that I don't think makes sense to put in the data model.)

-Osma

-- 
Osma Suominen
D.Sc. (Tech), Information Systems Specialist
National Library of Finland
P.O. Box 26 (Teollisuuskatu 23)
00014 HELSINGIN YLIOPISTO
Tel. +358 50 3199529
osma.suominen@helsinki.fi
http://www.nationallibrary.fi

Received on Tuesday, 15 April 2014 07:15:08 UTC