- From: Dan Brickley <danbri@w3.org>
- Date: Fri, 26 Aug 2005 09:47:41 +0100
- To: Mark van Assem <mark@cs.vu.nl>
- Cc: Antoine Isaac <Antoine.Isaac@kb.nl>, public-esw-thes@w3.org
Mark van Assem wrote: > > > Hi Antoine, all, > > The question of distinguishing between immediate parents/children is > also present for rdfs:subClassOf. It requires some calculation but > it's possible to determine the direct relationships (see also the > directSubclassOf operator in SeRQL [1]) by determining that there > isn't an intermediate class Z between classes X and Y. So a "direct" > relationship is not indispensible I think. > > Of course it would be easier if the relationships were there > explicitly from the start. If people think that's better to have then > I would consider putting skos:directBroader/Narrower into the core and > recommend its usage instead of putting it in the extensions [2], as > you already say. > > I couldn't come up with a use case in which it is *crucial* to make > the distinction, but maybe some others on this list can? I'm pretty wary of this notion of 'direct' versus implied, as it interacts awkwardly with the open-world assumptions behind Web (and Semantic Web) architecture... eg. class dan:Car subclass: mark:MotorCar subclass dan:ExpensiveMotorCar In the 'dan' scheme, we might assert that dan:ExpensiveMotorCar is a sub class of the class dan:Car. Elsewhere in the Web, someone else (eg. you) might describe a class 'MotorCar' that fits between the two in the hierarchy. Does this fact that someone has taken the time to document this new intermediate class make the relationship between dan:Car and dan:ExpensiveMotorCar any less direct? Perhaps only if the dan: schema is ammended to mention it? Even within a single vocabulary, we run into similiar issues around versioning, addition of new terms. "Directness" is relative. I expect query tools and databases (eg. built on SPARQL) will be used by applications who do care to know whether some relationship (eg. subClassOf) is directly asserted, versus implied, within a particular dataset. But I really have my doubts that this is usefully expressed in a de-contextualised way in terms of a 'directness' relationship between terms. These comments are mostly w.r.t. the general RDF issues, but I think they apply to directBroader too. Dan
Received on Friday, 26 August 2005 08:47:38 UTC