Re: Design Issues (1) - Specialised vocab vs. extensible modular voca bs?

Dan Brickley wrote:

 > Does this make sense?

Yup.

By the way the Gene Ontology is quite an interesting example in this 
respect. It has been migrating from a thesaurus like structure to a more 
ontology like structure. At the moment it has two link types, isa links 
which have strict subsumption semantics (as in RDFS et al) and part-of 
links which have less strict semantics. In the future they are 
contemplating spliting part-of into different flavours which could be given 
a stricter semantics.

In a meeting I was at recently it was suggested that the GO is easier to 
navigate than some of the deeper ontology-based domain models in similar 
domains because of this thesaurus-like heritage. The speculation was that a 
complex ontology is likely to benefit from a thesaurus-like navigation overlay.

Dave

> Yup, I think it is things "like" Thesauri that are in scope.
> 
> "like" being under-determined somewhat.
> 
> Here is my take:
> 
> RDFS/OWL are RDF's own way of doing this stuff. We don't need to 
> reproduce that.
> 
> SWAD-E's thes vocab is for representing KOS systems that don't (for
> whatever reason) directly map into RDFS/OWL, but do map into a 
> thesaurus-like network of relationships amongst named concepts.
> 
> So:
> 
> (1) out of scope
> fido --type--> Poodle --subClassOf--> Dog --subClassOf--> Mammal
> 
> (2) in scope
> 
> fido --bt--> Poodle --bt--> Dog --bt--> Mammal
> 
> 
> Obviously you can talk about the same stuff in both traditions.
> 
> The explicit RDFS/OWL view (ie. 1. above) is clearer and supports more 
> inference, eg.  you can deduce that fido is of type Mammal, thanks to
> RDFS's formal semantics.
> 
> (2) is more typical of what we see in the library world, where the 
> looser notion of 'bt' conflates a variety of distinctions.
> 
> So the idea here is to bootstrap the semantic web by allowing an 
> RDF representation of these fuzzier, messier but still useful 
> databases of related named concepts. Remodelling a thesaurus as an
> RDF vocabulary (RDFS/OWL) is expensive and time consuming. Dumping out 
> from a thesaurus into TIF should be easy and cheap, and allow some 
> benefit from use of generic RDF tools, although of course missing out 
> on other aspects of RDF which focus on the type hierarchies.
> 
> Does this make sense?
> 
> Dan
>

Received on Thursday, 6 November 2003 10:33:23 UTC