[MOBY-l] ontology

[Phillip Lord]

>>>>> "Simon" == Simon Twigger <simont at aa9pw.com> writes:


  Simon> As I got more idea of how Owl works, my thought was that the
  Simon> object ontology might be another branch of a moby
  Simon> 'uber-ontology' structure so that services could have an
  Simon> input and output properties which could be filled by classes
  Simon> from the object ontology branch. 

This certainly makes sense, and is effectively what we do within
mygrid. 


  Simon> By appropriately defining the service classes in the ontology
  Simon> in terms of input and output objects from the object ontology
  Simon> we could then reason over structure and generate a reasoned
  Simon> version that would put services in all the appropriate
  Simon> places, rather than the one place we currently use. I think
  Simon> this is more S-MOBY territory but would make service
  Simon> discovery more flexible and dynamic - users would register a
  Simon> service in terms of input and output objects (as they do now)
  Simon> and the service ontology would then classify the service into
  Simon> the appropriate type(s) for discovery.


Well, whether you need reasoning to do this or not depends on the
complexity of the definitions of types that you want. If you want to
allow fully OWL(-DL) expressive queries, then you will need
reasoning. If you want simpler things where, for instance, the inputs
and outputs are explicit named classes already in the ontology (rather
than OWL Class expressions) then you can probably make things
simpler. 

Either way, within mygrid, we have code base which may be of use in
helping to implement this sort of search backend. 



  Simon> The object branch in the service ontology file was my attempt
  Simon> to explore how this might work in OWL by defining input and
  Simon> output properties for services and seeing if I could fill
  Simon> them with objects and have the reasoner behave as expected. I
  Simon> should probably prune that from the public version and just
  Simon> work on classifying the services so things don't get unduly
  Simon> complicated at this stage. I think this is a good way to go
  Simon> but we should make some executive decisions balancing this
  Simon> type of development vs getting something useable in a
  Simon> reasonable timeframe.

Well, you need to be reasonably sure of your search technology (that
is the expressivity of your queries), if you are going to judge
usability. One of the reasons, for example, we stopped using reasoning
at run time in mygrid was the simple fact that we didn't have any cute
GUI's for forming complex queries...

Cheers

Phil