Re: SKOS Comment (Reference WD June 2008) - broaderTransitive < broader, narrowerTransitive < narrower

On Fri, 1 Aug 2008 at 18:52:25, Simon Spero <ses@unc.edu> wrote
>On Fri, Aug 1, 2008 at 4:44 PM, Leonard Will 
><L.Will@willpowerinfo.co.uk> wrote:
>
>>Instead of using the terms "broaderTransitive" and "broader", it would 
>>be clearer and in keeping with thesaurus practice to use 
>>"broaderGeneric", "broaderPartitive" and "broaderInstantial" as well 
>>as the general relationship "broader" (which would not be transitive 
>>if it allowed a mixture of the specific types).
>
>"The Hierarchical Relationship" is transitive across the subtypes 
>within the domain of Knowledge Organization Systems;  however the 
>subtype of relationship is lost.

Why then has SKOS introduced the idea that "broader" is not necessarily 
transitive, requiring the introduction of a different relationship 
"broaderTransitive"?

>BTI is considered by most to be a subtype of BTG, but obviously must be 
>the terminal node in a chain).
>
>car_wheel BTP car
>car BTG wheeled_vehicle
>wheeled_vehicle BTG vehicle
>-----------------------------------
>car_wheel BT car
>                 BT wheeled_vehicle
>                 BT vehicle
>car            BT wheeled_vehicle
>car            BTG vehicle
>wheeled_vehicle BT vehicle
>

Is it the case that in any indirect relationship like

car_wheel BT vehicle

inferred from a chain containing a mixed set of BTP and BTG 
relationships, the inferred relationship will be BTP? (It is true that 
car wheels are parts of vehicles and not kinds of vehicles). Can anyone 
think of a chain of relationships for which this is not true?

>Where BT is the broader relationship between subjects, as described by
>Svenonius, Soergel, Green, Milstead, Taylor, etc.    Soergel's treatment of
>pre-combination and poly-hiearachy is especially informative.

If he treats this specific issue of transitivity, a reference would be 
most appreciated.

>The BTG relationship allows additional inferences to be made about
>relationships between  classes of things that are described by those subject
>terms.  This  is the traditional subclass relationship.  Correctly declared
>BTG relationships can be used to infer corresponding OWL/RDFS ontologies in
>a fairly obvious way.  The mapping for BTI is also fairly clear.

Agreed.

>BTP can also be mapped somewhat simply, but possibly with a restricted
>semantics that require specialization.

The danger is that BTP can easily lead to wrong relationships, such as 
the incorrect

wheels BTP cars

because all wheels are not parts of cars. This is why BS8723-2 says

"8.3.3.1 The hierarchical whole-part relationship covers a limited range 
of situations in which the name of a part implies the name of its 
possessing whole. This applies to four main classes of terms: a) systems 
and organs of the body, b) geographical locations, c) disciplines or 
fields of discourse d) hierarchical social structures."

If we accept this limitation, mixed chains of BTP and BTG are unlikely 
to occur, so the problem of transitivity will seldom arise.

>If a current KOS defines a relationship  between terms where A *R* B and B *
>R* C , but not A *R* C,   *R* *doesn't* correspond to BT.
>By definition, the relationship is an associative one.  The correct way to
>map these relationships into SKOS is as a (possible subrole of)  *related*.

Are you saying here that if a relationship is not transitive, it should 
not be expressed as "broader"? I.e. that SKOS does not need to introduce 
"broaderTransitive" because "broader" is by definition transitive and if 
it is not the relationship should be associative rather than 
hierarchical?

I am inclined to agree with you, with the warnings above about the need 
to restrict the use of BTP. Perhaps the SKOS folk will come back to say 
why they disagree.

This is on the assumption that SKOS is modelling thesauri that conform 
to standards, which is what I think it should do. Some existing KOSs 
look like thesauri but have incorrect relationships - SKOS can encode 
these as though the relationships were correct, but users would have to 
realise that inferences cannot be drawn from them with any confidence.

The hierarchies of a classification scheme, which represent combination 
of concepts as well as just subordination of concepts, are not at 
present within the scope of SKOS, so it is inappropriate to try to use 
SKOS to model them until it is extended.

Leonard Will

-- 
Willpower Information       (Partners: Dr Leonard D Will, Sheena E Will)
Information Management Consultants              Tel: +44 (0)20 8372 0092
27 Calshot Way, Enfield, Middlesex EN2 7BQ, UK. Fax: +44 (0)870 051 7276
L.Will@Willpowerinfo.co.uk               Sheena.Will@Willpowerinfo.co.uk
---------------- <URL:http://www.willpowerinfo.co.uk/> -----------------

Received on Saturday, 2 August 2008 11:08:48 UTC