Re: SKOS transitive hierarchical relations

On 1/6/12 9:25 PM, Simon Spero wrote:
> On Fri, Jan 6, 2012 at 10:08 AM, Christophe Dupriez <christophe.dupriez@destin.be <mailto:christophe.dupriez@destin.be>> wrote:
>
>     I said that if you have ALL the possible broaderTransitive, then you should be able to sort the paths (remove the "shortcuts") and retrieve only the true simple "broaders".
>
>
> It depends on what you mean by /true/, /simple/, and /broader /; under closed domains, either explicit (i.e. if the set of concepts is explicitly closed by asserted axioms, or implicit (e.g. through completion under Negation As Failure); and if there are no redundant desired broader relationships, then
>
> (if (not
> (exists Z
> (known (and (broaderTransitive A Z) (broaderTransitive Z B)))))
> (broader A B))
>
> This is transparent to poly-hierarchy.
>
> This is similar to the semantics of broader matches , which are made to concepts in different concept schemes, which entail the non-existence of any terms in the target concept scheme which is narrower than the term to which the source term is related, but is still a broader match of the source term; that is, the broader match is a least upper bound.
>
> The problem with asserting broader and inferring broaderTransitive is that SKOS does not require that broader semantics be transitive. Thus, unless the vocabulary creator asserts broaderTransitive, it is impossible to determine whether the vocabulary creator believes it.


As it stands, the interpretation of broaderTransitive is "ancestor in the hierarchy", sort of. So if one asserts "A broaderTransitive B", one does *not* say that "B is the direct parent of A and I say that this relation is transitive" (as could be the case for a broaderGeneric inspired by ISO 2788). One says "B is an ancestor of A". That does not say much, and that's not something any vocabulary creator could not believe in: "ancestor", as less specific relation, is of course transitive.

Antoine

Received on Saturday, 7 January 2012 11:01:05 UTC