[Pellet-users] Rules with frequently changing ontologies

[Giovanni Clemente]

Hello,
I'm using Pellet to implement an access control system based on  
ontologies.
Answering an access request implies realization of a named individual  
to determine which classes
he is member of. It also implies the need to update the ABox (but not  
TBox) with new information
specific to each request.

I know that currently each ABox change needs a complete reload, and  
that incremental
inferencing will be soon available. Nonetheless, without rules,  
performance on small ontologies
are very good (far less than a second response time).

However, turning on rules causes an awkward performance impact: given  
an ontology composed of

about 20 classes
about 10 props
about 10 individuals (plus spurious ones caused by owlapi parsing of  
swrl rules)
about 10 subclassOf axioms
1 equivalentClass axiom

2 rules with P(x) head
1 rule with Q(x,y) head,

Pellet answers (correctly) in more than 330 secs on a P4-3GHz. I know  
that the first query to Pellet
causes an intensive precomputation, and that subsequent queries are  
much faster but,
given that I have to reload all ontologies on every request, this  
precomputation won't help
me much.

at http://pellet.owldl.com/ I read about the existence of 2  
strategies to cope with rules: I would know
which of them is best suited for my case and if it is possible to  
switch beetween them.

Thank you very much,
ciao
Giovanni


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.owldl.com/pipermail/pellet-users/attachments/20070312/ab5739e3/attachment.htm