Re: [Dbpedia-discussion] Using DBpedia resources as skos:Concepts?

Hello Simon & Simon,

>     Simon Spero wrote:
> 
>         There are practical implications for indexing that follow from
>         this decision.  For example, the SKOS broader relationship is
>         not transitive; this is hard to understand with a document based
>         domain of interpretation.  Without transitive BT relationships,
>         standard indexing behaviors like upward posting, or assigning
>         the most specific headings to a document are no longer possible
>          (or rather, give different results).
> 
> 
>     But that's what skos:broaderTransitive is for? As a super-property
>     of skos:broader it will always be inferred if you just declare a
>     skos:broader relationship between two concepts, plus it's
>     transitive. So if you want to find transitively broader concepts you
>     query for skos:broaderTransitive but if you want directly declared
>     relationships you just query for skos:broader.
> 
>     So not sure where your problem lies.
> 
> 
> BroaderTransitive is not supposed to be asserted,


Well, that's the convention. But in specific cases, if you want to assert it directly, you can of course. There is no formal constrain, only a practice that we wanted to highlight.


> and is not required to 
> be valid- “the semantics of skos:broader, [..] cannot enforce such 
> transitivity" (Issac and Summers 2009).  


That sentence is *not* meant to say that broaderTransitive's transitivity is not valid! We have a formal axiom that states that it is valid, and that's it.
The sentence was trying to say why in general skos:broader is not transitive. Ie., that it's not transitive in some schemes (yourself demonstrated that there were problems in LCSH, haven't you?).
But that being said, you can assert all the necessary triples for one scheme such that it's locally transitive in that scheme. Eg. you can assert A broader B, B broader C and also assert A broader C if you will.
The difference in the latter case is that the assertions result from a KOS publisher's explicit intention, endorsing the local transitivty behaviour. That's where the difference lie. broader is not transitive so that publishers who don't want to endorse transitivity won't be arm-twisted by SKOS semantics and inference engine. broaderTransitive is transitive so as to allow consumers of data to access the transitive closure of the hierarchy, even if that transitive closure was not asserted/endorsed in the first place.


> 
> The change to the semantics of broader was made due to confusion about 
> the meaning of transitive relationships that  spanned different  broader 
> relationships that arises if the document centric model is not 
> followed.  The example used is:
> 
> In many cases, hierarchical relations in a concept scheme can be 
> considered as transitive [OWL]. If |ex:animals| is broader than 
> |ex:mammals|, which is itself broader than |ex:cats|, it makes sense to 
> assert that |ex:animals| is broader than |ex:cats|. However, there are 
> "dirtier" hierarchies, especially in KOSs different from standard 
> well-designed thesauri, where such a feature would not be judged 
> appropriate. Consider for instance a case where |ex2:vehicles| is said 
> to be broader than |ex2:cars|, which is itself asserted to be broader 
> than |ex2:wheels|. It may be problematic if "wheels" is automatically 
> inferred to be a narrower concept with respect to "vehicles".  (Issac 
> and Summers 2009).
> 
> This is entirely unproblematic in KOSes, where the extensions of 
> Concepts are documents.


Even if one agrees with the extensions of concepts' being documents, then this is not automatically validating the transitivity assumptions, in many KOSs.
Imagine you have documents about very specifc types of religious building. Would you want them to appear under a very generic "religion" category? 
I was in a workshop on the UDC classification last week, and there was a talk where at some point the presenter made use of the transitivity assumption for retrieving documents. I was hidden deep in the audience, and I can tell you that some experts found that dangerous ...

Note that personally I'd be leaning toward the use of that assumption for retrieving documents. But again we cannot enforce that if there is no complete consensus. And in fact the optimal document retrieval solution probavly lies somewhere in-between, in the form of "controlled expansion strategies" such as the one Alistair investigated in his thesis [1]. And to do such controlled expansion, you need to be able to access skos:boader graphs that represent what is endorsed by the KOS publisher, not the transitive closure. And again, the users who want to use the transitivity assumption for document retrieval can of course ask for the broaderTransitive statements

Antoine

[1] http://purl.org/net/retrieval

Received on Friday, 6 November 2009 09:02:22 UTC