Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[protege-owl] SWRL evaluation in Protege

Martin O'Connor martin.oconnor at stanford.edu
Wed Apr 25 17:05:10 PDT 2007


> The Wiki page on SWRL evaluation 
> (http://protege.cim3.net/cgi-bin/wiki.pl?SWRLRuleEngineBridgeFAQ)  
> lists several limitations of the current implementation. 
>
> For example, "a ... property may be asserted for an individual as a 
> result of firing a SWRL rule, but a class level restriction in OWL may 
> forbid this property from belonging to that individual. As present, 
> these conflicts are not detected automatically, and resolving the 
> conflict - which is essentially between a SWRL rule and an OWL 
> restriction and has nothing to do with a particular rule engine - is 
> left to the user." 
>
> Can the possibility for such conflicts be predicted as rules are authored?

No - one would need to run a reasoner/inference engine to detect these 
inconsistencies. The exact same issue arises with OWL restrictions 
alone: one can easily write mutually contradictory restrictions when 
authoring an ontology - and only be able to detect these inconsistencies 
by running a reasoner.

>
> Also, "conflict-free execution of the classifier may also infer new 
> knowledge that may in turn produce information that may benefit from 
> further SWRL inference, a process that may require several iterations 
> before no new knowledge is generated."
>
> If the classifier and SWRL evaluation are run non-interactively, how 
> does one determine no new knowledge has been generated?

Both the SWRL bridge and reasoners that I am familiar with have API 
calls to determine if new assertions were made during a run.

As a practical issue, I am not sure that the lack of integration between 
a reasoner and inference engine is at all significant for the majority 
of ontologies. I have been using SWRL extensively in multiple projects 
for almost two years and have not run into a problem because of this 
disconnect and have not seen any emails on this mailing list reporting 
problems. When SWRL makes assertions, one can immediately run a 
classifier to check their validity. If they pass this test, they are 
guaranteed to be formally sound. That said, my goal is to have such an 
integrated solution in the SWRLTab in the not too distant future using 
the KAON2 reasoner/inference engine. I am awaiting the release of DIG 
2.0 to perform this integration.

Martin



More information about the protege-owl mailing list