Re: SKOS and Rules

On 27 Nov 2006, at 23:04, Antoine Isaac wrote:

>
> Hi Sean,
>
> You pointed at a very important problem here
>
>>
>>
>> The SKOS Core Guide includes reference to rules:
>>
>> http://www.w3.org/TR/swbp-skos-core-guide/#secaboutrules
>>
>> [[
>> Inference rules are part of SKOS Core. Inference rules are
>> described in prose, and where appropriate are expressed using the  
>> Jena
>> 2 rule syntax [Jena Inference], or as RDF statements using the OWL
>> vocabulary [OWL]. Inference rules appear in boxes such as:
>>
>> An example rule.
>>
>> (?x ex:p ?y)
>> ->
>> (?x ex:q ?y)
>>
>> prefix ex: <http://www.example.com/eg#>
>>
>> ]]
>>
>> I can see what's going on here, but it makes me a little
>> uncomfortable. If the SKOS recommendation is to make use of  
>> additional
>> infrastructure, shouldn't this be couched in terms of a standards
>> (e.g. RIF when it appears) rather than referring to some particular
>> implementation (Jena rules)?
>
> I completely agree with you. But I think that at the time when  
> these rules where specified standards where not easily available,  
> nor really helpful to get a simple message get through (RuleML/SWRL?)
>
>>
>> Is there going to be some notion of comformance? For example, does
>> this mean that I *cannot* provide a SKOS implementation without  
>> having
>> some RDF repository that implements the rules -- and in that vein,
>> what are the precise intended semantics of these rules?
>
> Alistair would perhaps be better placed than me to answer. However,  
> I have worked with SKOS for quite a while now, and to me it is now  
> quite natural to consider that the rules are part of the spec, and  
> therefore should be implemented if one wants to implement a proper  
> SKOS engine. This explain perhaps there have been cases of  
> vocabularies represented using SKOS, but no "official" SKOS  
> inference engine... (though such a thing wouldn't be too hard to  
> implement, I suppose)
> Concerning the semantics of the rules, well I suppose that the idea  
> was just to have some production rule that would enable to specify  
> formal inferences fitting what the SKOS model was trying to render.

I'm not a hardline DLista, but with my vaguely formal hat on, I'd  
certainly want to see this clarified. That would then make it easier  
to see if such an engine really wasn't "too hard to implement...."

> And I supposed that if OWL had offered means to represent general  
> relational composition, there wouldn't have been such Jena-like  
> axioms, but just OWL ones, as for e.g. the transitivity of  
> skos:broader.

I think the property chain inclusion axioms in the proposed OWL 1.1  
[1] would cover the subject generality rule. Not sure about the  
others though.

> Whatever be the case, it is clear that this should be clarified for  
> further version of SKOS. Notice by the way that some of the  
> concerned axioms (either OWL ones or "rules") are quite  
> controversial (well, I disagree with the relevance of some, and am  
> not the only one), and might disappear from a further version of SKOS.

Ok.

	Sean

[1] http://owl1_1.cs.manchester.ac.uk/overview.html

--
Sean Bechhofer
School of Computer Science
University of Manchester
sean.bechhofer@manchester.ac.uk
http://www.cs.manchester.ac.uk/people/bechhofer

Received on Tuesday, 28 November 2006 16:53:25 UTC