Re: Cross-ontologies reasoning

Hi Ugo,

This idea of making deductive logical inferences across ontologies is 
one of the principals of the OWL-DL flavor of the language.  You do it 
by establishing axioms that express equivalencies, sub-class, or other 
relationships between the two ontologies (or many more ontologies) and 
use a mechanism such as owl:import to provide a linkage.  If you have 
an inferencing technology, then you can maintain logical consistency 
across these relationships.  "Closeness" is a matter of interpretation 
and can be influenced somewhat by the form of the "bridge" axioms 
expressed.  If ontologies are far apart -- ie different concepts -- the 
logic processor would not infer that they represent the same or similar 
things.

Note also that in effect you can obtain a chaining effect from the 
ontologies --- the introduction of a 3rd ontology only requires 
bridging axioms to some of the concepts in 1 of the previous two 
ontologies, and so on.  So you are protected from an N-squared mapping 
problem.

The state of the art is that these systems are available.    The notion 
of federated or composite ontologies is directly instrumented in the 
OWL language.  Our company ships an OWL-DL suite that provides this 
capability, as well as authoring tools and aids that help to provide 
these "bridging" statements.

Hope this helps.

Jack
jack.berkowitz@networkinference.com

On 11 Dec 2003, at 17:20, Ugo Corda wrote:

> The recent discussions on semantics and UDDI made me wonder about what 
> happens when a query is formulated in terms of a particular ontology, 
> but the registry contains information related to ontologies other than 
> that one. I imagine some ontologies are so far apart (e.g. describing 
> completely separate vertical industries) that reasoning across them 
> would not make much sense. Other ontologies might be closer to each 
> other, and including them in the same reasoning process could make 
> more sense.
>
> Does anybody know what is the current state of the art in 
> cross-ontologies reasoning? Are there metrics that would allow us to 
> determine how closely related two separate ontologies are and how much 
> sense it would make to reason across them?
>
> Thank you,
> Ugo