Re: What do the ontologists want

From: Frank Manola <fmanola@mitre.org>
Subject: Re: What do the ontologists want
Date: Mon, 21 May 2001 13:08:56 -0400

> Peter--
> 
> It seems to me when you say "makes it essentially impossible to use the
> semantics for the positive ground triples to represent domain
> information" you're going a bit beyond what Pat said.  It seems to me
> that Pat was pointing out that all RDF allowed you to do in the example
> was define a collection of relationships between some instances
> ("positive ground triples");  it didn't give those relationships any
> special semantics, either logical (to support inferencing) or domain
> semantics.  If I say "author(a,b)", RDF doesn't define the semantics of
> what an "author" is either (we want DAML+OIL to do things like that). 
[
]
> However, I presumably can use a collection of "author(X,Y)" triples to
> represent useful domain information (given that my processor knows how
> to interpret "author" properly, of course).  This is what databases do
> all the time, and what the simplest examples of RDF illustrate.  

This is precisely the sort of thing that I meant when I said that there are
useful things that RDF can represent.  I'm sorry if I was unclear in what I said.

> Now suppose I introduce some new relationships like "premis" and
> "conclusion" (and possibly some other encoding tricks to get everything
> into triples).  It seems to me it's certainly possible to mix triples
> involving "premis" and "conclusion" (and any other relationships needed
> for the encoding) with triples involving "author", "subject", and other
> domain relationships, and not mess up whatever domain semantics I get
> out of the domain relationships, provided I can properly separate the
> stuff that is to be interpreted as part of the "more-expressive"
> formalism from the stuff that is to be the ground facts.  
> Or, to translate this into database terms, suppose I take my relational
> database and store my ground triples representing authors in it.  Now
> suppose I add some additional tables to the database to support an
> encoding of logical rules, to be operated on by some inference engine. 
> The same author information (domain semantics) can be extracted from the
> database as before those additional tables were added, so I haven't lost
> the ability to represent domain information (any ability you already
> had, that is).  But obviously you need to make sure that you keep the
> use of the domain tables appropriately separated from the tables
> representing the "more expressive formalism" if you're going to make any
> sense of what you have.  In other words, RDF doesn't restrict the
> relationships you define to any one level of abstraction (whether one of
> those levels is called "encoding" or not), but it doesn't provide any
> magic means of defining (much of) the semantics of those relationships
> (no matter what level they are), or sorting those levels out either.  

This is what I think that *cannot* be in RDF, at least as it stands now.
In RDF as it stands now the ``premis'' and ``conclusion'' triples are given
exactly the same semantic import as the ``author'' and ``subject''
triples.  Even worse, the triples that are used with the ``premis'' and
``conclusion'' triples are also given the same semantic standing, and these
triples will have ``author'' and ``subject'', or other domain predicates,
as their predicates.  There is no mechanism (in RDF) for separating the
triples that actually convey domain meaning from those that are just part
of implications, or negations, or existenials, or ....   

This is why I say that you cannot use RDF semantics when you have an
encoding of more-expressive formalism.  There is no way (in RDF) to
separate the triples that have direct semantic meaning from those that need
other treatment.

> --Frank

Peter F. Patel-Schneider
Bell Labs Research

> 
> "Peter F. Patel-Schneider" wrote:
> > 
> > In defense of stripped-down RDF, there is nothing technically wrong with a
> > logical formalism that can represent only positive ground triples.  Such a
> > formalism can certainly convey some useful semantic information.
> > 
> > It is just that such a representation formalism cannot be used to
> > *represent* anything more than positive ground triples.  Using positive
> > ground triples to encode a more-expressive formalism requires encoding, which
> > requires a new semantics, defined on top of the semantics for the positive
> > ground triples, and makes it essentially impossible to use the semantics
> > for the positive ground triples to represent domain information.
> > 
> > Peter F. Patel-Schneider
> > Bell Labs Research
> > 
> > From: Graham Klyne <GK@ninebynine.org>
> > Subject: Re: What do the ontologists want
> > Date: Mon, 21 May 2001 13:25:04 +0100
> > 
> > > I think this example+explanation is one of the clearer expositions of what
> > > RDF can and cannot do.  I suspect that many proponents of RDF recognize
> > > this without being very clear about how to articulate it.
> > >
> > > I don't think any of the long-term advocates of RDF expect it to be able to
> > > express all meanings (or even any universal truths?) using only semantics
> > > defined for core RDF.
> > >
> > > I think the interesting question is:  can the RDF core framework alone
> > > convey *any* useful semantics, or is it no more than an abstract syntax
> > > over which semantics must be defined?
> > >
> > > #g
> > >
> > >
> > > At 08:54 PM 5/18/01 -0500, pat hayes wrote:
> > > >>For an example, let me introduce a propositional logic and provide a
> > > >>rule R which says that given triple <a,b,c> anyone may infer triple
> > > >><d,e,f>.  This logic is not very expressive; it does not even allow
> > > >>conjunction in the premise:
> > > >>
> > > >>    <R, premise, RP>
> > > >>    <RP, subject, a>
> > > >>    <RP, predicate, b>
> > > >>    <RP, object, c>
> > > >>    <R, conclusion, RC>
> > > >>    <RC, subject, d>
> > > >>    <RC, predicate, e>
> > > >>    <RC, object, f>
> > > >>
> > > >>Each of these triples is true itself, while also building a structure
> > > >>for us.
> > > >
> > > >How does this convey the meaning that you indicate, ie that <d,e,f> can be
> > > >inferred from <a,b,c> ? It simply says that some things exist called 'R',
> > > >'RP' and 'RC', which stand in some undefined relationship to a, b, c, and
> > > >so on. The RDF data model provides no further meaning, and the model
> > > >theory for RDF provides no further meaning. So no inferences are sanctioned.
> > > >
> > > >If you want this kind of structure to actually mean somethingmore than
> > > >this - in particular, if you want it to have the force of an implication,
> > > >as indicated - then you need to state truth-conditions which support that
> > > >larger meaning. But those truth-conditions will have to refer not more
> > > >than  the RDF syntax; they depend on the particular relation symbols you
> > > >have used: in this case, 'premis' and 'conclusion'.  (You will also need
> > > >to relate <a,b,c> to the three triples with 'RP' in the subject, but I
> > > >presume that this wil be done by reification, so I won't dwell on it.) In
> > > >other words, you have now given those symbols a *logical* meaning: they
> > > >have become part of the logical syntax. This isn't RDF any more: it is
> > > >something else, implemented in RDF.
> > >
> > >
> > > ------------
> > > Graham Klyne
> > > GK@NineByNine.org
> > >
> 
> -- 
> Frank Manola                   The MITRE Corporation
> 202 Burlington Road, MS A345   Bedford, MA 01730-1420
> mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-8752
>

Received on Monday, 21 May 2001 13:56:01 UTC