RDF syntax for RIF

We have a requirement to tag rule sets with metadata indicating the 
"intended semantics (or semantic style)" of the ruleset.

In the charter (section 2.2.5) we have a requirement to consider 
expressing such metadata in RDF.

One approach to this (which has been mentioned at least in f2f 
discussions) is that we could have an RDF syntax for rules and rulesets, 
thus making it easy and natural to use RDF to provide the metadata for 
such rules. Metadata is, after all, one thing RDF is definitely good for.

I had wanted to follow up on this once the dust had settled on the next 
UCR round but some recent discussion on the W3C Semantic Web 
Coordination Group list have sufficiently reset my understanding of RIF 
it seemed better to raise it now.

I'd like to propose that RIF specify an RDF-based syntax for RIF rulesets.

In particular that we do something like:

Define RDFS or OWL classes:
   rif:Rule
   rif:RuleSet

and properties
   rif:head (domain rif:Rule, range rdf:XMLLiteral)
   rif:body (domain rif:Rule, range rdf:XMLLiteral)
   rif:rule (domain rif:RuleSet, range rif:Rule)
   rif:annotation (base for property hierarchy for RIF annotations)

So that we use the current proposed XML serialization of expressions but 
use RDF as a scaffolding to combine these, along with dialect tags, into 
rules and rulesets.

Such an approach:
  o is pretty simple
  o allows us to use RDF for tagging and thus also use SPARQL to
    query such annotations
  o allows RIF rules to be transmitted as RDF documents
  o allows RIF expressions to be parsed using standard XML tools
  o avoids issues of embedding RDF expressions directly as RDF statements
  o does not preclude having a pure XML syntax such for rulesets as well
    [though in that case I suggest we provide a GRDDL mapping from that
     to the RDF syntax so that the two can be mutually interchanged;
     indeed careful choice of the XML syntax could redender it parsable
     as if it were the RDF/XML syntax without modification :-)]

Is the *broad* thrust of this proposal reasonable?

If so then we can nitpick the details later.

Dave

Received on Monday, 25 September 2006 21:03:22 UTC