my notes from today's JC telecon on rule test cases and integration with OWL

From: Benjamin Grosof (bgrosof@MIT.EDU)
Date: 07/01/03

  • Next message: Harold Boley: "Equality Hubs"
    % notes from JC telecon 6/24/03
    % by Benjamin Grosof
    o rule test cases:
    Mike has a bunch of wish cases, e.g., about equivalence
    e.g., using sameIndividualAs or cardinality-1
    he's working on an engine that has some knowledge of OWL, and uses N3 syntax,
    that can do some of the rule test cases
    todo:  talk with Mike about integrating his ideas/prototype with the
    SweetOnto/KAON work on DLP (Description Logic Programs) which was demo'd
    by Boris Motik at WWW-2003 semantic web developers day
    Benj:  wrt style:  suggestion:  more prose comments,
    organize by point made, e.g., domesticflight-1.n3 and uncle-1.n3
    and siblingfather-1.n3 all give examples of ontological knowledge
    that are hard/impossible to express in OWL-DL
    "<>" denotes the current KB
    discussion about whether
    equality should be treated explicitly, and defined
    by rules,
    vs. built into the language/engine
    Benj:  we obviously want the capability for the latter,
    but recommend the former as one available compliance level in the beginning,
    for flexibility, since
    many practical rule systems do not have it built in already with the full
    strength of FOL semantics
    Mike: this issue is very related to whether our requirement for the rule
    language is
    to extend OWL
    to work optionally with OWL
    Mike's approach builds in knowledge of OWL to the rule language and system
    Benj:  let's list this as an objective
    Benj:  SweetOnto/KAON implementation that we have right now first translates
    OWL/DL stuff into LP, then does inferencing in an LP rule system;
    but it's nice to have the OWL syntax (say, for DLP expressive subset)
    Benj during discussion with Pat:
    nonmon about equality, say default distinctness via NAF,
    is not well studied in the KR literature,
    so it would be more cautious (wrt sticking to what's already
    well understood) to define equality as an ordinary
    predicate and then explicitly axiomatize its substitutivity, transitivity,
    Pat:  but can't we rely on NAF as a mechanism to define nonmon aspects
    of equality?
    Benj:  yes, that's the direction I would take
    Benj:  wrt more test cases beyond DLP:
    - for ontological knowledge, e.g., in MIT Process Handbook or
    various frame-based systems, often want nonmonotonic inheritance, thus
    beyond DLP
    - it's pretty easy to give examples of cardinality or disjunction or
    existentials that go expressively beyond the current version of DLP
    Pat in discussion with Mike and Benj:
    let's aim to have some OWL syntax directly within the rule language
    Benj:  yes, this is a large motivation of the DLP work,
    but this might be beyond our V1 because it involves more complexity
    consensus:  yes; need RDF syntax for the rule language, and a corresponding
    human presentation syntax e.g. cf. N3
    Pat:  wrt syntaxes, let's look at TRIPLE by Stefan Decker and Michael Sintek
    Prof. Benjamin Grosof
    Web Technologies for E-Commerce, Business Policies, E-Contracting, Rules, 
    XML, Agents, Semantic Web Services
    MIT Sloan School of Management, Information Technology group or

    This archive was generated by hypermail 2.1.4 : 07/01/03 EST