I am using Protege 3.4 and I built my ontology and extended it with SWRL rules using SWRL and sqwrl built ins . The rules classifies instances of a class of the ontology ; the RHS is class assertion of an instance .It works successfully but when I change the values of the causes in the LHS no reclassification is done the instance is asserted to the same class
ex: SWRL rules:
Message(?m) hasInterest(?m,?i) hasCategory(?m,?c) sqwrl:makeset(?s1,?i) sqwrl:makeset(?s2,?c) sqwrl:intersection(?s3,?s1,?s2) sqwrl:size(?n,?s3) swrlb:greaterThan(?n,0) -> Ham(?m)
Message(?m) hasInterest(?m,?i) hasCategory(?m,?c) sqwrl:makeset(?s1,?i) sqwrl:makeset(?s2,?c) sqwrl:difference(?s3,?s1,?s2) sqwrl:size(?n,?s3) swrlb:greaterThan(?n,0) -> Spam(?m)
So once the message instance (m1) is classified as ham for example as i= sports and c=sports , whenever I change the values of i= movies ( interests) for the message instance (m1) it will always be ham . I understood that this is because the class type is asserted . So my question is Why does this happen ? How to reclassify instances as I need a dynamic way for message classification
I editted the question with my work's example ,what do you think ?
For some reason, even changing the classes , the first rule is still true and messge keeps being classified as Ham. Can you share your ontology? I could test it on Protege and try to figure out the issue. One question: are you restarting the reasoner after changing the values to have all the rules reevaluated?
Yes I start the process of mapping the asserted facts ; classes and instances and the rules( so the fact that m1 is ham is mapped as the type of the individual already to the engine ) and then I run the Jess engine and then map the result back to the ontology . Ok I will share it with you . When I searched I found that this is because of non monotonic reasoning of OWL and SWRL and that the SWRl rule facts are asserted not inferred so I posted the question to check and find an alternative such that the asserted fact are inferred .
What do you mean by "reclassify "? Like removing the instance from one class and attaching it to another?
Yes moving the instance ( Fact assertion , RHS) from a Ham class to Spam class according to the change in the causes ( LHS).
If this can not be done in SWRL rules what rule language do you propose instead to have a non monotonic inference.
You could create a new ontology, that imports your ontology. When you run the SWRL rules in this ontology, all new assertions will be stated in the importing ontology, while your imported ontology will be unchanged. If you want to "discard" your inferences, you just create a brand new ontology that imports your main ontology.
I did not try this, but theoretically it should work. At least in Protege 5.2.
Ideally all the inferences created by SWRL rules should be presented as inferred knowledge, and not asserted in the ontology by default, but this is a limitation of the SWRLTab. Unfortunately improving this behavior would take too much effort, and the Protege team does not have the time to fix it at the moment. You can send such questions to the Protege user mailing list (after you subscribe), where you may get prompt and precise answers from the SWRLTab author himslef: http://protege.stanford.edu/support.php
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators