Named graphs & SNAF -- is there more to it?

The rule language workshop (1) must have been a great event.
The subsequent discussion on SNAF vs NAF (2) falls within
recurring issues of open vs. closed world semantics as argued in (3).

These issues are clearly important; however, it seems to me that
we lack a consensus for specifying what context is and how we can
use context for limiting the scope of existential and universal 
quantification.

At the level of practical APIs and tools, we could use the notion of
a named graph to reify the intension of a context (graph name) and
the extension of that context (graph contents). Unfortunately, it seems
to me that named graphs don't quite fit well within the semantic web
stack of (3): RDF, RDFS, OWL, SWRL, FOL


In SPARQL, one could write a precise interpretation
of the query from (3) about "asking if pat knows exactly one person" 
like this:

Given:

:context1 {  <#pat> <#knows> <#jo> }

Query:

SELECT DISTINCT ?person WHERE { GRAPH :context1 { <#pat> <#knows> 
?person } }

Named graphs seem to me the right way to go to make subtle but important 
distinctions
among several notions of context. In practice, it is rare to find 
notions of context explicitly
represented. This makes some ontologies particularly ambiguous when the 
ontology itself
behaves implicitly as the scope of quantification for the ontology's 
axioms. For ontologies
constructed via embedding, like DOLCE, it could lead to errors if a 
quantifier, ?x, could
bind to an individual and the reification of that individual as two 
possible solution bindings
when they are semantically the same thing modulo reification.

-- Nicolas.

(1) http://lists.w3.org/Archives/Public/public-sws-ig/2005Jun/0008.html
(2) 
http://lists.w3.org/Archives/Public/public-rule-workshop-discuss/2005Jun/0028.html
(3) 
http://www.cs.man.ac.uk/~horrocks/Publications/download/2005/HPPH05.pdf 
<http://www.cs.man.ac.uk/%7Ehorrocks/Publications/download/2005/HPPH05.pdf>

Received on Wednesday, 6 July 2005 20:53:12 UTC