RDF-LP [was Re: Expressing operators and functions in RDF]

Drew,

Sorry for the delay of my response, I was doing my homework last week by
reviewing logic basics (in particular FOPL).
By doing so, I now understand better the underlying rationale of your
DRS approach and I'm now convinced that it is the right way to go.
I have started to craft my own ontology in OWL called RDF-LP based on
your ideas with DRS, Common Logic, OWL Rule Language,OpenCyc, SUO and my
reading on the FOPL. I wish to post the current state of my thinking on
a wiki web site for discussion. Do you think ESQ Wiki
http://esw.w3.org/topic/FrontPage appropriate for such a discussion ?.
Suggestions are welcome.

The goal of this ontology is to define a logic framework in RDF to
EXPRESS, VALIDATE and EVALUATE any formula based on FOP. OWL Rule
Language would be an application schema of the framework.

Here the big lines of my approach:

First layer is a mathematical foundations based on relations to put in
common properties of FOP:

- Define the concept of Relation with its subclasses
FixedArityRelation,Unary,Binary.Ternary, Quaternary, Quintary,
VariableArity Relations.
- Define convenience properties of FixedArityRelation:
argument1,argument2,argument3,...argument5 (OpenCyc and SUO don't go
beyond 5 arguments) and arguments (for variable arity).
- Define Relation metataproperties  such as arity, argument1Isa,
argument2Isa,..,argument5Isa, argumentsAre, resultIsa that help for the
validation of formula. 
- Define TruthFunction with its subclasses
Predicate,UnaryPredicate,BinaryPredicate,...QuintaryPredicate
- Define Function with its subclasses
UnaryFunction,BinaryFunction,...QuintaryFunction,VariableArityFunction.

The second layer based on the first layer defines FOP terminology.

- Term, Variable, VariableList, TermList, Formula and its subclass
AtomicFormula (predicate), NonAtomicFormula (function).
- LogicalConnective (subclass of Term) and its subclasses And,Or,Implies
(Rule), Iff,Not
- Quantifier: ForAll, ThereExists, ThereExistAtLeast,ThereExistAtMost,
ThereExistExactly. 

I have the following questions: 

- Should we have Predicate and Function as argument0 of a formula or
should we have Predicate and Function subclass of AtomicFormula and
NonAtomicFormula respectively (subject of the statement) ?
- How N-Ary predicate are bridged with RDF triples ? Do we assume the
subject of the triple is argument1, and object of the triple are lists
with arg2,arg2,...) ?
  What about unary predicate in RDF ? Do we have often this case except
OWL:Class ? 

Do you think this approach is viable and compatible with the semantic
web vision, Common Logic and DRS ? 

Best regards
 
Stephane Fellah

Received on Monday, 17 November 2003 10:56:02 UTC