Re: RDF core: abstract syntax and semantics - an attempt

At 02:49 PM 6/21/01 -0500, you wrote:
>In the attached document, you write:
>
>>At first sight, [an RDF graph] may appear to be simply 'true' or 
>>'false'.  In the absence of reification, this is enough.
>
>Please excuse my naivete, but what does it mean for an RDF graph to appear 
>'true' or 'false'.

A nice question to let me introduce my thinking some more...

The truth or falsity of an RDF graph is in the domain of 
interpretation:  when interpreted it either corresponds to some state of 
affairs (truth) or does not (falsity).

A (the?) domain of interpretation that particularly interests us contains 
the Web as we know it.  Consider the following RDF graphs interpreted with 
respect to the Web:

   <http://www.aaronsw.com/> dc:author "Aaron" .

   <http://www.aaronsw.com/> dc:author "Graham" .

The interpretation of <http://www.aaronsw.com/> is your web page, dc:author 
is some function that associates A and B with 'true' if they are related as 
indicated by the English statement "B is the author of A", "Aaron" is a 
label understood to refer to you, and "Graham" is a label understood to 
refer to me.

Then, we can say that under this interpretation the RDF graph containing 
the first statement can be  seen as representing a truth, and the graph 
containing the second statement as representing a falsity.  This is what I 
mean by appearing 'true' or 'false'.

...

But I have also claimed that this is not sufficient to capture the meaning 
of an RDF graph.

Suppose I have two subgraphs g1 and g2:

g1=   ex:stmt ex:saidBy "Graham" .

and

g2=   ex:stmt rdf:type rdf:Statement ;
               rdf:predicate dc:author ;
               rdf:subject <http://www.aaronsw.com/> ;
               rdf:object "Aaron" .

These two graphs are each well-formed expressions of RDF.  But from these 
we can also create a 3rd graph g (using the 'g ::= g1 g2' abstract syntax 
production):

g=    ex:stmt ex:saidBy "Graham" .
       ex:stmt rdf:type rdf:Statement ;
               rdf:predicate dc:author ;
               rdf:subject <http://www.aaronsw.com/> ;
               rdf:object "Aaron" .

The predicate 'ex:saidBy' is to be interpreted as the relation between A 
and B that is true if A and B are related as in "B said A".

Now, this combined graph tells us additional information about the 
statement said by Graham.  For example, Graham has said that Aaron is the 
author of something.  This additional information must come from the 
interpretation of "g2".  If the interpretation is simply 'true' or 'false', 
that cannot convey the kind of additional information suggested 
above.  Hence, I conclude that the interpretation of an RDF graph must 
include not only truth or falsity, but also information about statements 
that are contained in the graph but not asserted (if RDF is to be capable 
of expressing ideas like the example above, which RDFM&S seems to be quite 
clear *is* a goal of RDF).

...

Does this help to explain what I am trying to do here?

#g


------------------------------------------------------------
Graham Klyne                    Baltimore Technologies
Strategic Research              Content Security Group
<Graham.Klyne@Baltimore.com>    <http://www.mimesweeper.com>
                                 <http://www.baltimore.com>
------------------------------------------------------------

Received on Friday, 22 June 2001 08:34:05 UTC