[Pellet-users] Tracing the Cause of a Lengthy Inference

[Michael Smith]

On Tue, 2008-04-29 at 15:22 -0400, JPQ wrote:

> I have an ontology that imports a second ontology, which is in the
> local workspace. The namespace URI in the <owl:imports> statement
> matches the base URI in the ontology to be imported. Pellet is run
> over the ontology that imports the second ontology. Unfortunately, it
> runs for a long period of time. The imported ontology inferences in
> about nine minutes. The ontology that imports the second ontology adds
> approximately one dozen classes and two dozen properties. I don't
> understand what is preventing this ontology from inferencing within a
> similar time frame.

The types of axioms in the importing ontology are much more important
that the number of classes and properties.  Without seeing the actual
ontologies in question, we can't explain much.
 
> One additional point of interest. The bulk of the ontology is
> comprised of a third ontology imported into the second ontology, yet
> the second ontology inferences while the ontology in question does
> not.

Same thing.  Imports structure really helpful because we have no idea
how the axioms in each ontology interact without seeing the ontologies.

> When I manually Pellet, no error is returned.

Pellet from the command line?  Which options are you using?  How long
does the run take?

> Lastly, Pellet is running within TopBraid Composer.

> At this point, I'm baffled as to why this occurs. Aside from checking
> that the namespaces match as described earlier, is there some
> prescribed method of tracing this problem?

If you are running Pellet from the command line, you can pass the
-timing option and get some information about where Pellet is spending
its CPU time.  If you'd like to report those results back here, maybe
there will be something telling.  I suggest, however, finding a way to
send the ontologies.

Regards,
-- 
Mike Smith

Clark & Parsia