On 23/6/09 09:33, Martin Hepp (UniBW) wrote:
> Hi Dan:
> I think Alan already gave examples this morning. An ontology can contain
> statements about the relationship between conceptual elements - classes,
> properties, individuals - that (1) influence the result to queries but
> (2) are not likely retrieved when you just dereference an element from
> that ontology. The more complex an ontology is, the more difficult is it
> to properly modularize it.

Indeed, I missed his mail until after I'd sent mine. And the examples 
are helpful. However they are - for the non-SemWeb enthusiast - 
incredibly abstract:


What I'd love to see is some flesh on these bones: a wiki page that 
works through these cases in terms of a recognisable example. Products, 
people, documents, employees, access control, diseases, music, whatever. 
I want something I can point to that says "this is why it is important 
to take care of the formalisms...", "this is what we can do so that 
simple-minded but predictable machines do the hard work instead of us...".

> But basically my main point is that the use of owl:imports is defined
> pretty well in
> and there is no need to deviate from the spec just for the matter of gut
> feeling and annoyance about the past dominance of DL research in the
> field. And as the spec says - with a proper owl:imports statement, any
> application can decide if and what part of the imported ontologies are
> being included to the local model for the task at hand.

+1 on respecting the specs, but also all know that not every piece of 
specification finds itself useful in practice. Having a worked-through 
to the instance level account of why owl:imports is useful would help. 
There is no compulsion re standards here: if someone is happy publishing 
RDFS, we can't make them use OWL. If they're happy using OWL we can't 
make them use RIF. If they're happy with RIF 1, we can't make them use 
RIF 2 etc. Or any particular chapter or verse of those specs.

What we can do is ground our evangelism in practical examples. And for 
those to be compelling, they can't solely be at the level of properties 
of properties; we need an account in terms of instance level use cases too.



Received on Tuesday, 23 June 2009 08:54:51 UTC