Re: SKOS transitive hierarchical relations

Thanks for the explanation about causes and effects Antoine, it does makes sense

> Well, nothing prevents you from doing that, but then you have lost some information. Despite the property names, asserting a skos:broader is indeed *more precise* than asserting a skos:broaderTransitive. Because the former implies the latter (mathematically speaking, the skos:broaderTransitive is the transitive closure of the skos:broader relation, http://en.wikipedia.org/wiki/Transitive_closure)

So as I understand it, broader/narrower is a safe basis for everyone, and then one could augment the detail level with transitive statements.

It's not explicit in the SKOS primer, I understood that I could use transitive relations instead of B/N to express a hierarchical tree where all relations are transitive (animals > mammals > cats, all cats are animals)

I saw another example recently of this "all-transitive hierarchical tree" [1], the author also seems to model its vocabulary only based on transitive relations [2]


> Well, the optimal scenario is that you would assert your skos:broader "individual steps in the hierarchy" first:
> ex:ConceptA skos:broader ex:ConceptB
> ex:ConceptB skos:broader ex:ConceptC
> 
> And then your thesaurus app could create and maintain the transitive relation:
> ex:ConceptA skos:broaderTransitive ex:ConceptB
> ex:ConceptB skos:broaderTransitive ex:ConceptC
> ex:ConceptA skos:broaderTransitive ex:ConceptC

ok, the app could detect some multi-steps relations and create transitive links when they exists
but in your example, the two first statements are not really useful, are they ?

My goal is to be able to do query expansion, looking for things linked to a concept, but also things linked to a skos:broader concept, and finally things linked to a skos:broaderTransitive concept




[1] http://156.35.31.156/pscs/page/naics/2012/resource/4529
[2] http://156.35.31.156/web/en/softwate-and-datasets/datasets/38-pscs

--
Dominique Guardiola, QUINODE
• http://www.quinode.fr/
• Tel : 04.27.86.84.37
• Mob : 06.15.13.22.27

Received on Thursday, 5 January 2012 08:48:41 UTC