[219]                               home                           [221]


Saturday, November 19, 2005


 The BCNGroup Beadgames

National Project à 

Challenge Problem  à

 Center of Excellence Proposal à





Discussion at ONTAC forum

ONTAC stands for Ontology and Taxonomy Coordinating Working Group

It is a working group of

Semantic Interoperability Community of Practice (SICoP)



Communication from Ralph Hodgson, of Top Quadrant




I have just caught up with the thread on "Some thoughts on hub ontology and merging sources" and because I am unaware of how to post a response to an archive (I don't think it can be done) this note is posted as a new thread.


In response to John Sowa's posting "Re: [ontac-forum] Some thoughts on hub ontology and merging sources" at http://colab.cim3.net/forum/ontac-forum/2005-11/msg00040.html. To the piece:

"Merging and integration are required to achieve tight coupling,
but not every system requires tight coupling.  Loose coupling is
much easier to build and maintain than tight coupling.    (013)"

We agree whole-heartedly. Our work with ontology-based systems has always required modularity of approach.  FEA-RMO was done this way. Because of the modularity we have been able to use FEA-RMO with a DODAF ontology. The SRM of FEA-RMO we have used independently.  In our work with NASA, the ontologies we have been working on are also small and diverse across a number of disciplines. At the University of Texas Medical Center, the SAPPHIRE ("Situation-Awareness and Prevention of Public Health Incidents using Reasoning Engines") ontologies are also decoupled, in fact, used within a mediation architecture. For these reasons we have created an ontology for experessing the dependencies between ontologies. This is called OARS, which stands for "Ontology Architecture Requirements Specification".


In response to Paul Prueitt's posting "RE: [ontac-forum] Some thoughts on hub ontology and merging sources" at http://colab.cim3.net/forum/ontac-forum/2005-11/msg00034.html. To the piece:

"The white paper introduces the OWL standardization process as if completed
and completely satisfactory.  But this hides the fact that Protégé is incomplete, and I
assume that Top Quadrant used Protege to encode the FEA RMO? (016)"

FEA RMO was indeed developed in Protege 2000, although certain Protege limitations present at that time required us to manually create and maintain one of the 'bridging' ontologies.  In fact, a longer version of the white paper we have submitted to GSA included some reflections on the state of the tooling. The work has been done about 10 moths ago now. Protege has moved on (at least partially) and some of the problems are now fixed.


We made sure that the model was 100% percent standard compliant with the current OWL-DL spec. If the paper gave an impression that the standardization was fully complete, it was not intentional. I just came back from the 2005 ISWC and can report that a lot of interesting things are happening with rules (including a change to the semantic web layer cake), further development of OWL (for example, to support composition statements) and with semantic web services. As standardization of OWL (and rules) continues, see http://www.w3.org/2005/Talks/1110-iswc-tbl/#(12) for Tim Berners-Lee's talk at ISWC2005,  we look forward to ways of expressing modular ontologies with the possibility of selective imports.


One of the most forward-looking aspects of our current work is in pluggable architecture for reasoning engines. The goal is to be able to use different types of reasoning over a given knowledgebase combining, for example, probabilistic (Bayesian) reasoning with rule-based reasoning.


One thing that has changed in the last year, which we are glad of, is that a semantic solution no longer has to be a 'cutting edge', research project. Many projects can be (and are being) done using standards and tools 'as is'. They've gotten to the point where they are good enough to deliver value.


One of our current projects is, for example, building a production, ontology-based system to match job requirements with a database of available candidates. The idea is not to remove the human beings from this process, but to assist them by producing a higher quality candidate 'short lists'. Another project has to do with integrating ontologies with an enterprise search portal. We are also doing a data interoperability project where ontologies are used to translate outputs of different tools. All of these projects are possible with standards 'as is', but they will definitely benefit from improvements and advancement in the standards.