[Pellet-users] explanation

[Evren Sirin]

On 6/12/08 4:04 AM, Hongkai Liu wrote:
> Hello,
>
> I have been using Pellet to give an explanation of inconsistent ABox.
> I found out the ABox assertion given by Pellet does not syntactically
> equal to the one in the input ABox. 

By not syntactically equivalent you mean that the ordering of elements 
in constructs like "and" and "or" are different that the one in the 
input yes you are right that Pellet will not return the same ordering. 
This is same for OWLAPI too, for example, which stores the concepts in 
an intersection and union as sets and the ordering of elements in the 
set is not relevant.

The other thing I see to be syntactically different is the 
representation of one-of construct for which Pellet has no internal 
representation (internally it is represented as a disjunction of 
nominals). If you retrieve explanations through OWLAPI we map the 
internal representation of Pellet to corresponding OWL constructs which 
will give you syntactically equivalent results (again syntactic equality 
here assumes set equality). There is no similar mechanism for KRSS. What 
you can try is use the KRSS interface of OWLAPI to load the your KRSS 
ontology and use the OWLAPI interface to get explanations. There is 
still a small chance that explanations will be syntactically different 
than originals (especially for the kind of complex, synthetic concepts 
as below) but it should be better than the direct KRSS interface.

> Here is the input:
>
> (DEFINE-PRIMITIVE-ROLE R_SPOUSE)
> (DEFINE-PRIMITIVE-ROLE R_HASP)
> (INSTANCE IND_JOHN (OR (NOT (AND (OR HAPPY (OR (ONE-OF IND_MARY)
> (ONE-OF IND_JOHN))) (NOT (NOT TOP)))) (AND (OR (OR (ONE-OF IND_MARY)
> (ONE-OF IND_JOHN)) (ALL R_SPOUSE (OR (NOT (AND (OR HAPPY (OR (ONE-OF
> IND_MARY) (ONE-OF IND_JOHN))) (NOT (NOT TOP)))) (NOT (ONE-OF
> IND_MARY))))) (ALL R_SPOUSE (OR (OR (ONE-OF IND_MARY) (ONE-OF
> IND_JOHN)) (OR (NOT (AND (OR HAPPY (OR (ONE-OF IND_MARY) (ONE-OF
> IND_JOHN))) (NOT (NOT TOP)))) (NOT (ONE-OF IND_MARY))))) (AND (OR (NOT
> (ONE-OF IND_JOHN)) (ALL R_SPOUSE (OR (NOT (ONE-OF IND_JOHN)) (OR (NOT
> (AND (OR HAPPY (OR (ONE-OF IND_MARY) (ONE-OF IND_JOHN))) (NOT (NOT
> TOP)))) (NOT (ONE-OF IND_MARY)))))) (OR (NOT (ONE-OF IND_JOHN)) (ALL
> R_SPOUSE (OR (NOT (ONE-OF IND_MARY)) (OR (NOT (AND (OR HAPPY (OR
> (ONE-OF IND_MARY) (ONE-OF IND_JOHN))) (NOT (NOT TOP)))) (NOT (ONE-OF
> IND_MARY)))))) (OR (NOT (ONE-OF IND_MARY)) (ALL R_SPOUSE (OR (NOT
> (ONE-OF IND_JOHN)) (OR (NOT (AND (OR HAPPY (OR (ONE-OF IND_MARY)
> (ONE-OF IND_JOHN))) (NOT (NOT TOP)))) (NOT (ONE-OF IND_MARY)))))) (OR
> (NOT (ONE-OF IND_MARY)) (ALL R_SPOUSE (OR (NOT (ONE-OF IND_MARY)) (OR
> (NOT (AND (OR HAPPY (OR (ONE-OF IND_MARY) (ONE-OF IND_JOHN))) (NOT
> (NOT TOP)))) (NOT (ONE-OF IND_MARY))))))) TOP)))
> (RELATED IND_JOHN IND_MARY R_SPOUSE)
>
> -------------------------
> I call Pellet with the following method:
>
> public void consistent (String Name, boolean UNA) {
> 		if (UNA)
> 			PelletOptions.USE_UNIQUE_NAME_ASSUMPTION = true;
> 		else
> 			PelletOptions.USE_UNIQUE_NAME_ASSUMPTION = false;
> 		PelletOptions.USE_TRACING = true;
>         KnowledgeBase kb = new KnowledgeBase();
>         kb.setDoExplanation( true );
>         KRSSLoader loader = new KRSSLoader();
>         try{
>         loader.load( Name , kb );
>         }catch (Exception e){
> 			System.err.println("Error: " + e.getMessage());
> 		}
>         kb.prepare();
>         //System.out.println( "Expressivity: " + kb.getExpressivity() );
>
>         boolean consistent = kb.isConsistent();
>
>         System.out.println( "Consistent  : " + (consistent?"Yes":"No") );
>
>         if(!consistent) {
>         	Set<aterm.ATermAppl> exp = kb.getExplanationSet();
>         	myPrinter.printATermAxioms(exp);
>         }
>         //*/
> 		
> 	}
>
> --------------------
> Here is the output of
>
>  WARN [main] (KnowledgeBase.java:1573) - Inconsistent ontology.
> Reason: Individual IND_JOHN is forced to belong to class
> and([not(value(IND_JOHN)),not(HAPPY),not(value(IND_MARY))]) and its
> complement
>  WARN [main] (KnowledgeBase.java:1575) - ExplanationSet:
> [type(IND_JOHN,or([and([_TOP_,and([or([all(R_SPOUSE,or([or([not(value(IND_MARY)),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(IND_MARY)]),HAPPY])]))]),not(value(IND_MARY))])),not(value(IND_MARY))]),or([all(R_SPOUSE,or([or([not(value(IND_MARY)),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(IND_MARY)]),HAPPY])]))]),not(value(IND_JOHN))])),not(value(IND_MARY))]),or([all(R_SPOUSE,or([or([not(value(IND_MARY)),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(IND_MARY)]),HAPPY])]))]),not(value(IND_MARY))])),not(value(IND_JOHN))]),or([all(R_SPOUSE,or([or([not(value(IND_MARY)),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(IND_MARY)]),HAPPY])]))]),not(value(IND_JOHN))])),not(value(IND_JOHN))])]),all(R_SPOUSE,or([or([not(value(IND_MARY)),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(IND_MARY)]),HAPPY])]))]),or([value(IND_JOHN),value(IND_MARY)])])),or([all(R_SPOUSE,or([not(value(IND_MARY)),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(I
>  ND_MARY)]),HAPPY])]))])),or([value(IND_JOHN),value(IND_MARY)])])]),not(and([not(not(_TOP_)),or([or([value(IND_JOHN),value(IND_MARY)]),HAPPY])]))])),
> prop(R_SPOUSE,IND_JOHN,IND_MARY)]
>
> The first assertion pinpointed by Pellet (1.5.1) is not any one in the
> input ABox. I did not check weather it is equivalent semantically to
> some one in the input. I knew that Pellet does pinpointing with "not
> top" instead of "bottom". I would like to let Pellet give me the
> explanantion which is a subset of the original ABox. Changing bottom
> to not top is ok since I can modify the input. However, it seems that
> Pellet also changes other things. Is that true? Is there any global
> variable setting this up?
>   

No, your best chance is to use OWLAPI interface as I described above.

Cheers,
Evren
> Best regards.
>
> Hongkai Liu
> _______________________________________________
> Pellet-users mailing list
> Pellet-users at lists.owldl.com
> http://lists.owldl.com/mailman/listinfo/pellet-users
> _______________________________________________
>
> Sponsored by Clark & Parsia, LLC http://clarkparsia.com/
>