RE: ACTION: task force unasserted triples

A few more observations about the owl:Dark proposal 
(http://lists.w3.org/Archives/Public/www-webont-wg/2002Apr/0290.html).

Given an RDF graph G, consider the subgraph containing all the 
triples in G that do not contain any uriref that occurs as the 
subject of a triple of the form

aaa rdf:type owl:Dark .

which is rdfs-entailed by G, ie which is in the rdfs-closure of G. 
Call this the 'light' subgraph of G, light(G).  Any RDF graph has a 
unique light subgraph which can be figured out by an RDFS reasoner.

(This allows implicit 'darkening', eg by defining an rdfs:subClassOf 
owl:Dark and asserting something to be in the subclass, or by using 
rdfs:range. An alternative proposal would be to require the triple to 
occur in the graph explicitly. While syntactically simpler, that 
would complicate the relationship between RDFS and OWL since drawing 
a valid conclusion in RDFS could alter the 'darkness' of the 
vocabulary. )

The desired semantic relationship between the languages can then be 
stated as the condition: Any satisfying OWL-interpretation of G is a 
satisfying rdfs-interpretation of light(G).

Lemma: If S rdfs-entails G, then light(S) rdfs-entails light(G).

This means that drawing valid RDFS conclusions from some OWL (even on 
the 'dark' vocabulary) isn't going to produce any unwanted 
conclusions. It might produce some conclusions that are dark, but it 
isn't going to produce any light ones that it shouldn't produce, ie 
that OWL would find embarrassing.

More later, hopefully with some test cases.

Pat

-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Wednesday, 24 April 2002 22:22:28 UTC