[MOBY-dev] Layering OWL on top of BioMOBY RDF descriptions

[Duncan Hull]

Gary Schiltz wrote:

>I can then paste that RDF into the converter at http://www.semanticmoby.org/developer/dev-tools.jsp
>  
>
...and thanks Gary too...

Actually, while we're on the subject of RDF...BioMOBY has done a great 
job of demonstrating some of the strengths and limitations of using RDF 
to represent bioinformatics service metadata.

The next layer in the semantic web cake [1] is to add OWL descriptions. 
Semantic MOBY, BioMOBY and myGrid have all made various moves in this 
direction. With OWL, we get more expressivity and the ability to reason 
over our services, in particular we possibly get more "bang for our 
buck", specifically:

1. We can distinguish between classes and instances (individuals)
2. Properties won't be able to have properties any more (something 
RDF/RDFS lets you do)
3. We can distinguish between language constructors and ontology vocabulary
4. We have localised range and domain constraints
5. We get existence/cardinality constraints
6. We get  transitive, inverse and symmetrical properties
7. We get better reasoning support using "native" reasoners

(see [2,3] for details)

So, which bits of this rich service metadata are going to be
a) realist to capture?
b) useful for biologists and bioinformaticians describing and finding 
services?
c) useful for registry providers to manage their services?

I realise these are all very open questions, but I'd be interested to 
know what moby-devvers think...

Duncan

[1] http://www.cs.man.ac.uk/~horrocks/Publications/download/2005/HPPH05.pdf
[2] 
http://www.cs.man.ac.uk/~horrocks/Teaching/cs646/Slides/ontologies.pdf 
(see problems with RDF/RDFS)
[3] http://www.cs.man.ac.uk/~horrocks/Publications/download/2003/HoPH03a.pdf

-- 
Duncan Hull
http://www.cs.man.ac.uk/~hulld/
Phone: +44 (0) 161 275 0677