Re: [RIF] Extensible Design: Horn semantics and syntax... RDF/OWL integration conventions

On Thu, 2006-09-14 at 13:23 -0400, Boley, Harold wrote:
> Michael & I have completed our Horn semantics and syntax
> actions from Tuesday, as part of the RIF Extensible Design:
> http://www.w3.org/2005/rules/wg/wiki/B._Extension%3A_RIF_Rule_Language
> http://www.w3.org/2005/rules/wg/wiki/B.1_Horn_Rules

The code I wrote to translate N3-rules also translates RDF/XML.
So I can run it on the OWL tests... for example, attached
find ,statecode.xml, the output of running it on...
  http://www.w3.org/2002/03owlt/I5.1/premises001 
and ,list.xml, the output of running it on
  http://www.w3.org/2002/03owlt/I5.5/conclusions005

The RDF parser in swap/cwm treats turns first/rest triples into
something more like a function term. So that's the way I have
written them out in RIF:

    <Atom>
      <Rel>holds</Rel>
      <Ind iri="http://www.w3.org/1999/02/22-rdf-syntax-ns#type" />
      <Expr>
        <Fun>list</Fun>
        <Ind iri="http://www.w3.org/2002/03owlt/I5.5/premises005#a" />
      </Expr>
      <Ind iri="http://www.w3.org/1999/02/22-rdf-syntax-ns#List" />
    </Atom>


Specifying a list function symbol and the relationship to first/rest
involves induction, I suppose. I think it's worth doing, though.


My approach to RDF "triples" is to use a 3-place "holds" relation:

    <Atom>
      <Rel>holds</Rel>
      <Ind iri="http://example.org/vocab#stateCode" />
      <Var>g3</Var>
      <Data>KS</Data>
    </Atom>

This allows us to write rules corresponding to the RDFS semantics
and parts of OWL, as in the owl pD* paper
  http://www.websemanticsjournal.org/ps/pub/2005-15

I transcribed those rules into N3 and then played
around with them a bit...
  http://www.w3.org/2000/10/swap/util/owlth.n3

Jos has done something similar...
  http://www.agfa.com/w3c/euler/rpo-rules.n3 

Attached find ,rpo.xml , the output of my N3-to-RIF
translator on that rpo-rules.n3.

For example
  {?P rdfs:domain ?C. ?S ?P ?O} => {?S a ?C}.

turns into...

  <Implies>
    <And>
      <Atom>
        <Rel>holds</Rel>
        <Ind iri="http://www.w3.org/2000/01/rdf-schema#domain" />
        <Var>P</Var>
        <Var>C</Var>
      </Atom>
      <Atom>
        <Rel>holds</Rel>
        <Var>P</Var>
        <Var>S</Var>
        <Var>O</Var>
      </Atom>
    </And>
    <Atom>
      <Rel>holds</Rel>
      <Ind iri="http://www.w3.org/1999/02/22-rdf-syntax-ns#type" />
      <Var>S</Var>
      <Var>C</Var>
    </Atom>
  </Implies>


FYI, I have a couple versions of RDFS semantics in rules;
one transcribed faithfully from the spec...
  http://www.w3.org/2000/10/swap/util/rdfs-rules.n3 
and one that omits "boring" rules like "everything is a Resource"...
  http://www.w3.org/2000/10/swap/util/rdfs-nice.n3

and for those who want to play along at home, this is the
exact command line I'm using:

~/w3ccvs/WWW/2000/10/swap$ PYTHONPATH=.. python2.4 n3absyn.py
http://www.w3.org/2002/03owlt/I5.1/premises001 --rif |tidy -i -xml
>,statecode.xml

and this assumes you have swap checked out, per
 http://www.w3.org/2000/10/swap/doc/cwm#dev

And if you want to use tidy...
 http://tidy.sourceforge.net/


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E