Re: Using rdf reification to nest statements in N3 like contexts

From: "Graham Klyne" <GK@NineByNine.org>

> The main thrust of my note [1] was to try and codify, in terms of
> relationships between the interpretations in a "context structure",
> simple patterns of contexts in a way that could lead to some simple
> some lifting rules.  Although I've hinted at some patterns, they are
> far from being nailed down and I think there will always be the
> possibility of additional patterns:  I don't expect a single definitive
> answer for every possible use of contexts.

I know.  But here, I though we were just talking about the syntax that would
allow us to represent assertions in contexts; and in particular how you
proposed to project those representations into RDF triples in section 6 of
your note below.

http://www.ninebynine.org/RDFNotes/UsingContextsWithRDF.html#xtocid-63039726

My point is that If we use your proposed syntax, then we are severely
limited because the tuples in any context are dependant on the level of
nesting.  If we want to talk about a inner nest, we can only refer to it
using that same tree nesting.

For example, let me assert something about the inner most assertion in your
test case: {C says D}.

  I believe {C says D}.

In your proposal that expression will cause 'C says D' to make one RDF
reification quad and will assert another triple between 'I' and that
statement.  Right?

But back to your example expresseion which asserts:

  A believe { B says {C says D} . B a liar } .

Here your proposal will doubly reify 'C says D'.  The doubly reified 'C says
D' will not test equal to the singly reified 'C says D' even though it is
the same exact content.  I don't think the content of a context is dependant
on the level of its nesting.  Rather the logical relationship *between*
contexts is what is dependant on the level of nesting.

> I'm sorry, but the mentograph didn't "speak to me" (I don't know how
> to interpret it), so I can't answer that.  I prefer to try and use (albeit
> imperfectly) the language that is used more generally in the field of
> KR, as that helps me to leverage some work that has been done.

You mean like KIF.  Thing is that mentography is not a language, rather it
is just grammar; it is pure uninterperted  syntax.  It's just labeled
directed graphs.  It carries with it no commitment whatsoever to any
ontology, logical primitives, or semantics.   But the claim remains
unchallenged that we can project any language whatsoever into this grammar,
it's corresponding graphical notation, and\or database structure.  We can
project RDF, N3, KIF, DAML, OWL, just as easily as we can project English.

Please see the new mentograph:
http://robustai.net/mentography/mentography_example_1.gif

When I introduced context into the diagrams I naturally started combining
labeled directed graphs with Venn diagrams.  There is some precedence for
this see "Projections in Venn-Euler Diagrams" :

http://www.it.bton.ac.uk/research/vmg/papers/Projections.pdf

But the mentograph below, which I made to grok your proposal,  was just
simple RDF triples.  You should have been able to read the RDF Ntriple
syntax directly off the graph: one arrow - one Ntriple.   If you can't
interpret it, then you can't interpret Ntriples.   The question remains: is
this what you are proposing?

http://robustai.net/mentography/context_n3_rdf_liar_comparison1.gif

Seth Russell

Received on Sunday, 17 March 2002 20:36:37 UTC