Re: Expressiveness of RDF as Rule Conclusion Language (was Re: W hat is an RDF Query? )

>From: "Pat Hayes" <phayes@ai.uwf.edu>
>
>>  >  Drew McDermott said:
>>  >
>>  >     Okay, if you change Sandro's "document scope"
>>  >     to "graph scope," you're right.  But then a rule like
>>  >
>>  >      (R1 ?x ?y) |-   (R2 ?y ?x)
>>  >
>>  >     becomes unstatable, because the ?x and ?y in
>>  >     the graph on the left are different variables from the
>>  >     ?x and ?y in the graph on the right.
>>  >
>>  >     Unless I'm missing something.
>>  >
>>  >Well, If I understand RDF graphs, I think you are.  Don't forget every
>term
>>  >(variable or not) has to be a URI in a RDF graph.
>>
>>  No! Blank nodes and literals are not URIs in an RDF grpah. The
>>  distinction is important for just this reason. bNode identifiers in
>>  an N-triples document are local to the document and do not have
>>  global scope.
>
>Which is why I used the word 'term' ... I guess I should have used the word
>'vocabulary' directly from the MT.  One practical way to make a schema that
>extended RDF to express Drew's formula, above, would be to invent URIs for
>the variables.

Oh, I see. But URI's aren't like variables, because they have global 
scope. Blank nodes are more like variables, which is why I jumped on 
you. That is, RDf graphs do have things in them that are *exactly* 
like existential variables in logic, but they are not URIs, and 
mustn't be made into URIs.

>  My only point was that those URI would force the left ?x to
>be the same identical node as the right ?x.
>
>The mentograph shows that clearly:
>http://robustai.net/mentography/entailsIdentity.gif   Incidentally in
>mentographs blank  nodes (unnamed nodes) show up as boxes with nothing
>inside the box where the name of the box goes. I love simple things.
>
>>  >  But if we scope formulas
>>  >to a context (a collection of statements) the ?x ?y in the left hand of
>the
>>  >formula become *the same* as the ?x ?y on the right .... that's the way
>RDF
>>  >works .... doesn't it ?
>>
>>  Nope, because RDF doesn't have contexts. (There are things like this
>>  in N3, but then N3 goes beyond RDF.)
>
>Alas ... but I'm sure it will someday ... aren't you?
>
>Actually I think a case could be made that we can do contexts in the current
>RDF since it  allows bags of statements:

Bags, but not of statements (unless you reify; but then you aren't 
asserting the things in the bag).

>substitute the word bag for the
>word context and use the MT to interpret a bag of statements as a set of
>statements.  Now I'm sure I didn't say that correctly, but if you blur your
>eyes a bit, maybe you will see what I meant.  I think our abilities to use
>RDF for logical inferences is severely diminished without contexts ... don't
>you?

Yup. But then RDF is very good for what it does, so lets leave it to 
do that and use something else to do other things, right?

>  How do we say this set of statements is monotonic and this set is not?
>
>>  The graph-merging rules described in section 3 of the RDF MT document
>>  should make this clear: if you merge two RDF graphs then you *must*
>>  merge nodes with the same URI, but you *must not* merge blank nodes.
>
>Why would (must?) one interpret a variable as an anonymous node?

See the MT document, section 2.

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 Saturday, 13 October 2001 00:49:47 UTC