Re: [SKOS] the return of transitive and subproperty (was Re: SKOS comment: change of namespace (ISSUE-117))

Hi Bernard,

You are correct, of course.  If I ignore the semantics implicit in SKOS 
tags and just treat them as pointers then it doesn't really matter 
whether the attribute is transitive or not.

But to me, the argument is about the primary use case of SKOS.  If SKOS 
is first and foremost designed as a thesaurus interchange language, then 
it's essential that <skos:broader> be non-transitive (see the 
polyhierarchy example I gave to Simon for an illustration of why).

My goal is avoid (a) seeing a substantial portion of the SKOS user base 
using properties in a way that undermines the semantic intent of the OWL 
properties; and/or (b) hurting uptake by forcing people to use 
less-than-optimal and non-intuitive designators such as 
<skos:minimalBroader>.

Cheers,

-- Stephen.

Bernard Vatant wrote:
> 
> Stephen
> 
> Your argument makes sense of course. Some remarks, nevertheless, still 
> in defense of my proposal
> 
> Stephen Bounds a écrit :
>>
>> One other observation which may or may not be accurate from the point 
>> of view of the broader SKOS community.
>>
>> I've always seen SKOS as being primarily a tool for vocabulary 
>> representation, *not* as a universal semantic reasoning tool.
>>
>> In other words, the goal of SKOS is to to accurately represent a 
>> thesaurus-like hierarchy in an easily understood manner using a common 
>> XML dialect.
>>
>> So I think having to do something like this:
>>
>> > 1. Declare only the direct broader assertions
>> > 3. Compute the transitive closure of broader
>> > 4. Compute the minimalBroader for each concept
>> > 5. Check if 4. matches 1.
>>
>> just to establish a parent-child relationship between nodes in the 
>> hierarchy greatly compromises the effectiveness of SKOS as a simple, 
>> expressive, declarative language.
> Actually, you don'*have* to do that. Most of the time, you can take it 
> for granted that the asserted broader *are* the minimal broader. So 
> simple applications can rely on this simple heuristic. Only direct 
> broader being asserted, your application can handle this in a very basic 
> way, without applying transitivity, and you have the same results as in 
> the current semantics (broader = minimal broader).
> You'll have to go through computing if you want to check vocabulary 
> consistency, apply semantic extension, whatever.
>>
>> At the moment in SKOS, I can establish the parents and children of a 
>> node by doing a simple XPath test on <skos:broader> and 
>> <skos:narrower>.  I don't want to lose that by reverting 
>> <skos:broader> back to a transitive property.
> You still can do that in most cases, if you don't compute the transitive 
> closure. Just take what is asserted ...
> 
> Bernard
> 
> 

Received on Thursday, 31 July 2008 14:05:59 UTC