Re: [TED] An alternative proposal for the technical design

Michael Kifer wrote:
> 
>>From your proposal:
> 
>     The parties in the interchange of a RIF document (that is, of rules
>     serialized as a RIF document) are assumed to have agreed on the meaning of
>     atomic expressions, that is, each party knows how to interpret them in
>     their own rule language,
> 
> Is it a set of bilateral agreements or a semantics? If you really mean to
> have a semantics then why don't you specify it and explain how exactly it
> differs from the other proposals. Right now your proposed semantics is no more 
> than handwaving.

A set of bi- (or multi-)lateral agreements, for all practical purposes. 
I am not sure that it differs from the other proposals, though.

>      The meaning of a ruleset is that all the rules have to be evaluated one at
>      a time in some order until none needs or can be evaluated anymore
>      (typically, because their consequent is satisfied or their antecedent is
>      not; but it could also be otherwise, depending on the application: in
>      application 2 above, each rule needs be evaluated only once)
> 
> Is that your "semantics"?

For a ruleset, yes. The semantics for a rule being: "the consequent 
expression is expected to be satisfied for all of the bindings produced 
by the initial quantification expression that satisfy the antecedent 
expression".

I did not call it a semantics for the same reasons that made you add 
quotes, but I guess it is a semantics nonetheless.

>      The rather vague "semantics" proposed above for rules and rulesets seems to
>      be able, for instance, to preserve the meaning of a rule/ruleset across a
>      large class of general logic programmes with a stable model semantics and a
>      large class of rulesets with the usual operational semantics of production
>      rule languages.
> 
> Stable model semantics, really? How?

Well, I may have been a bit careless, here...

What I meant is that, if you translate a ruleset from RIF into a LP 
language with a stable model semantics, what you can infer from it 
(given a set of facts) is (seems intuitively to be) compatible with the 
"rather vague semantics" described above. Same if you translate it into 
a PR language with the "usual operational semantics".

So, if you know that your ruleset is compatible with that "rather vague 
semantics" of my RIF proposal, and you want to publish it, you can just 
translate it into that RIF and leave it to the retriever to translate it 
back into whatever they want. It will not necessarily guarantee that all 
the parties will infer exactly the same "facts", but they will all be 
compatible with your "intended meaning" of the ruleset. Or, at least, 
that is the idea.

Now, why "stable models"? The (maybe poor) thinking behind is that 
binding all the variables produces a "ground equivalent" of the ruleset 
in the evaluation context. Evaluating each rule one at a time in some order:
- the consequent of rules in the antecedent of which there is a conjunct 
that is the negation of an atomic expression which is satisfied in the 
evaluation context will not be considered: these rules can be deleted 
without changing the outcome of the evaluation of the ruleset;
- the conjuncts, in the antecedent of a rule, that are the negation of 
atomic expressions that are not satisfied in the evaluation context do 
not impact the evaluation of the rule: they can be deleted without 
changing the outcome of the evaluation of the ruleset.

If the evaluation context is a stable model of the ruleset, that 
evaluation procedure should not modify it, even if you "enforce" the 
consequent.

My point is not to say that this must be the "semantics" of a ruleset in 
RIF. Rather, I meant to suggest that something as simple and vague could 
still be useful for the purpose of rule interchange.

Does this clarify?

Christian

Received on Tuesday, 31 October 2006 19:35:24 UTC