Re: Logic Layering Issues for DAML Rules

From: Ian Horrocks ([email protected])
Date: 03/18/03

  • Next message: Eric Prud'hommeaux: "Re: notes from our working meeting today on query and rules stuff"
    On March 16, Peter F. Patel-Schneider writes:
    > From: Sandro Hawke <[email protected]>
    > Subject: Logic Layering Issues for DAML Rules
    > Date: Sat, 08 Mar 2003 07:51:16 -0500
    > 
    > > 
    > > We've talked about the DAML Rules language, like RuleML, having at
    > > least one of its concrete syntaxes be RDF-Graph-based like OWL.  This
    > > raises concerns over logic layering; we need to handle universally
    > > quantified variables and some forms of negation.  (An aside: when
    > > people say "Horn rules" are they thinking only definite clauses, or
    > > full Horn clauses, with headless rules, giving us classical negation?
    > > My understanding is that's not as widely implemented, but since I'm
    > > tempermentally inclined towards a full FOL syntax, I view it as a step
    > > in the right direction.)
    > > 
    > > There were various comments about this at SWArch [1] yesterday, and I
    > > think I heard Pat and Ian say they knew how to do it, and maybe even
    > > saw Peter nod, but I didn't hear how.  
    > 
    > Maybe you saw my head jerk from astonishment, but certainly not a semantic
    > nod.  :-)
    > 
    > > When I've suggested doing it by
    > > encoding the syntactic structures of the rules into RDF and then using
    > > a limitted truth predicate (as in KIF [2]) to indicate which such
    > > structures are intended to be asserted as rules, people start to look
    > > very concerned.  
    > 
    > Truth predicates are an extremely powerful tool, but one that requires
    > great care or else serious problems will result.  
    > 
    > The problems are not limited to paradoxes.  Suppose you (by accident)
    > closed off the universe of non-empty predicates, say by 
    > 
    > 	sh:Rule rdf:subClassof _:x .
    > 	_:x owl:onProperty sh:ruleHead .
    > 	_:x owl:allValuesFrom _:y .
    > 	_:y owl:onProperty sh:predicate .
    > 	_:y owl:allValuesFrom [ list of predicates ] .
    > 
    > What is then the meaning of a rule that is missing a predicate?
    > 
    > > In any case, do we have a solution at hand, or is this going to be a
    > > major obstacle?
    > 
    > Major obstacle, provided that you want semantic layering to work.
    > 
    > If you 
    > 1/ want to only use RDF syntax, and
    > 2/ want the syntax to have its RDF meaning,
    > you are going to have problems.  The problems may be surmountable, with
    > great effort, but do you want to expend this effort, particularly when
    > relaxing either constraint above makes the problems largely go away?
    
    Agreed. As far as I am concerned, "knowing how to do it" meant knowing
    how to write FOL syntax down using RDF (which doesn't seem too
    difficult). Semantic layering issues are another thing
    altogether. Given the difficulties we had making this work for OWL
    (kind of), I believe that we should give up on it altogether for more
    expressive languages.
    
    Ian
    
    > 
    > >    -- sandro
    > 
    > Peter F. Patel-Schneider
    > Bell Labs Research
    > Lucent Technologies
    > 
    


    This archive was generated by hypermail 2.1.4 : 03/18/03 EST