W3C home > Mailing lists > Public > www-archive@w3.org > October 2002

Consensus on semantic layering

From: Frank van Harmelen <Frank.van.Harmelen@cs.vu.nl>
Date: Sun, 08 Sep 2002 23:11:07 +0100
To: Jim Hendler <hendler@cs.umd.edu>, Guus Schreiber <guus@swi.psy.uva.nl>, Dan Connolly <connolly@w3.org>
Cc: www-archive@w3.org
Message-ID: <250709944.1031526667@dhcp-w3c-209.hpl.hp.com>

Below are the notes on which the WebOnt WG agreed in a straw poll concerning 
the semantics for OWL.

Frank.
   ----

Consensus on Semantics for OWL
------------------------------

SYNTAX
------

- Large OWL does not impose any restrictions on how the OWL vocabulary
  can be combined (as long as they form legal RDF)

- Fast OWL assigns meaning to a subset of this
  Characterisation 1 of Fast OWL:
    any RDF graph generated from the abstract syntax via its proposed
    translation to RDF
  Characterisation 2 of Fast OWL:
    any set of RDF triple that conform to the restrictions in Pat/section 6
    - names are separated into classes/objects/properties
    - logical vocabulary is not applied to logical vocabulary

Unclear whether these two characterisations actually coincide
or whether another triple-based characterisation is needed

Any Fast OWL RDF graph is a Large OWL RDF graph, but not vice versa
Intuition: Fast OWL \approx DAML+OIL
Intuition: Fast OWL does not permit to mix OWL contents with arbitrary
           RDF contents

SEMANTICS
---------
- Three styles of defining semantics:
  - native FOL-style
    rdf:type = unary predicate application
    john type person = person(john)
        rdfs:subClassOf = implication between unary predicates
     person subClassOf mamal = person(X) -> mamal(X)
    rdf:Property = binary predicate
    union/intersection/complement = or/and/not
  - SKIF-style
    allows quasi-higher order expressions
    rdfs:Class rdf:type rdfs:Class  = Class(Class)

[
  - situation-calculus style:
    rdf(O,A,V) + full axiomatization
    e.g. need to axiomatised transitivity of subClassOf
    Dan's work, axiomatic DAML+OIL semantics
]

- native FOL-style semantics has been given for Fast OWL
- SKIF-style semantics can be given for Large OWL
- on Fast OWL, these two model theories hopefully correspond:
  if KB+C is in Fast OWL, then

  KB large-OWL-entails C  iff  KB Fast-OWL-entails C

Question: precise characterisation of requirements on KB+C for this to hold

- if the two happen not to agree, then Fast-OWL-entails is normative

MODUS OPERANDI
--------------
- Given a knowledge-base, tool1 must support the tight semantics
  while tool2 may support the liberal semantics
- On any expression in Fast OWL, these two would agree,
  For any expression outside Fast OWL,
  Tool1's behaviour would be undefined,
  while tool2 would return answer according to liberal semantics.
Received on Tuesday, 8 October 2002 06:14:54 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 7 November 2012 14:17:23 GMT