Re: A proposal for a unitary RIF phase 1

From: Chris Welty <cawelty@frontiernet.net>
Subject: Re: A proposal for a unitary RIF phase 1
Date: Mon, 29 May 2006 09:06:49 -0400

> Peter F. Patel-Schneider wrote:
> >     A Proposal for a Unitary Language for RIF Phase 1
> 
> There are many communities in the RIF WG, I'm not sure everyone knows, 
> or has the same definition for, what unitary means.  Do you intend here 
> anything more than RIF I will not admit infinite sentences?

[See separate response.]

> > Peter F. Patel-Schneider
> >
> > Overview:
> >
> > 1/ The language of RIF Phase 1 is function-free Horn clauses.  There
> >    is a human-readable, functional-style syntax, plus an XML
> >    interchange syntax.
> 
> Why function free?

No particular reason, except that neither RDF nor OWL have functions.

> > 2/ RIF Phase 1 includes predicates for the reasonable XML Schema
> >    datatypes, plus various built-in predicates over these datatypes.
> >
> > 3/ A RIF Phase 1 knowledge base is a set of RIF Phase 1 documents plus
> >    an optional set of OWL DL documents closed under OWL imports.
> >
> > 4/ The meaning of a RIF Phase 1 knowledge base is given by a standard
> >    model-theoretical semantics.
> >
> > 5/ Compliance for formalism X will be determined by the presence of a
> >    non-trivial subset of X that can be mapped into RIF Phase 1
> >    knowledge bases in a deduction-preserving mapping.

[...]

> > SEMANTICS
> >
> > An interpretation is an extension of an OWL DL interpretation
> > (http://www.w3.org/TR/owl-semantics/direct.html), extending the
> > vocabulary to include RIF relation names and mapping them to tuples
> > over R union LV.  RIF relation names that are OWL DL class names,
> > datatype names, individual-valued property names, data-valued property
> > names, or annotation property names have the same extension as given
> > by the appropriate part of the OWL semantics.
> 
> The OWL-DL semantics segregates the relation domain by arity, requiring 
> that unary relations be disjoint from binary.  Does this restriction 
> extend out the arity space, or is there just one additional set of 
> non-OWL-relations that is disjoint, or is the restriction removed, or....?

The proposal does not currently have any such requirements on rule
predicates.  You could even use OWL class names in trinay atoms or OWL
individual-valued property names with a data value as the second argument,
but such atoms would always be false.

> > A RIF rule is true in a RIF interpretation precisely when every
> > mapping from variables in the rule into R union LV that makes each
> > atom in the antecedent of the rule true in the obvious extension of
> > the OWL DL semantics also makes the consequent in the rule true in the
> > same way.
> >
> >
> > COMPLIANCE
> >
> > Compliance for formalism X will be measured as follows.
> > a/ Partial mappings will be provided between the syntaxes of X and RIF
> >    Phase 1, including mappings between X's data language and OWL DL.
> > b/ A subset of RIF Phase 1 will be identified as being X-compliant.
> > c/ For that subset the deductive behaviour of X must mirror reasoning
> >    in the RIF Phase 1 in the sense that ground consequences for
> >    knowledge bases in this subset are the same for RIF Phase 1 and its
> >    mapping into the syntax of X.
> 
> There is some overlap between your RIF I and OWL DL, that is there are 
> things that can now be said in two ways, the most obvious of which is 
> subClassOf.  Is the mapping responsible for generating one or the other, 
> or does the semantics "know" about it, or ... ?

The mappings do not have to work in any particular way.  I have decided
preferences in which way the mappings should work, but this is not
reflected in this proposal.

> > Compliance of a rule system with RIF Phase 1 will be defined as follows.
> > 1/ The formalism underlying the rule system must be RIF Phase 1
> >    compliant as defined above.
> > 2/ There must be a tool that implements the syntax mapping.
> > 3/ There must be a comprehensive set of RIF knowledge bases for which
> >    the equivalence of deductive behaviour has been reasonably 
> > demonstrated.
> 
> -- 
> Dr. Christopher A. Welty                    IBM Watson Research Center
> +1.914.784.7055                             19 Skyline Dr.
> cawelty@frontiernet.net                     Hawthorne, NY 10532
> http://www.research.ibm.com/people/w/welty

Peter F. Patel-Schneider

Received on Tuesday, 30 May 2006 09:26:38 UTC