RE: Rule qualifications and different rule types

Adrian - just catching up on some emails... I think for RIF to cover
"CEP" we would need to have the concepts of "event" and "time" covered -
which might be a challenge ... 


Paul Vincent 
+1 650 206 2493 / mobile +44 781 493 7229 


> -----Original Message-----
> From: Adrian Paschke [mailto:adrian.paschke@gmx.de]
> Sent: 03 March 2009 15:58
> To: 'Gary Hallmark'; 'Christian de Sainte Marie'
> Cc: Paul Vincent; 'RIF WG'; 'Serrano-Morales, Carlos A'; 'Berlioz-
> Matignon, Carole Ann'
> Subject: Rule qualifications and different rule types
> 
> Hi All,
> 
> With respect to my
> 
> ACTION: Adrian to investigate use cases where a specific Rule
construct
> would be needed (in future dialects, e.g. for CEP rules), as an
> indicator
> for whether and how urgently such a construct would be needed in PRD
> 
> 
> If RIF should be a general interchange format we
> 
> 1. need to support different types of rules
> 2. need to support different rule qualifications such as priority
> values or
> temporal constraints such as validity times or fuzzy, uncertainty
> quantifications, etc.
> 
> 
> With respect to 1:
> 
> There are many other rule families.
> 
> For instance, defeasible logic which distinguish between defeasible
> rules,
> where conclusions can be "defeated" by other rules with higher
priority
> and
> strict rules, which are like standard "if-then" derivation rules.
> 
> For instance, reactive rules which add an explicit event part, i.e.
"on
> Event if Condition then do Action".
> 
> Currently, we have Implies for if-then production rules and if-then
> derivation rules. Instead of introducing many other specialized rule
> construct we could generalize the Implies construct and reuse it in
> these
> rule families, e.g. for defeasible rules there would be an additional
> attribute on Implies indicating if the rule is strict or defeasible.
> For
> reaction rules, we would introduce an event part.
> 
> However the Implies construct itself is semantically misleading since
a
> reaction rule is not an implication rule. So, probably a more general
> "Rule"
> construct would make sense.
> 
> 
> 
> With respect to 2:
> 
> Qualifications are needed in various ways. We need them on rule sets,
> i.e.
> on the complete Group. But we also need them on the rule level.
Implies
> could be used, i.e. we could add qualification (e.g. as attributes or
> subelements) to the Implies. However, then unconditional actions and
> facts
> would need to be wrapped by the Implies construct. Again, calling a
> fact and
> implication is semantically incorrect - so a general "Rule" construct
> would
> be the solution, since a fact is an unconditional rule.
> 
> 
> -Adrian
>

Received on Sunday, 26 April 2009 02:26:59 UTC