uos-convene
[Top] [All Lists]

Re: [uos-convene] What needs done here + technical feasability

To: Upper Ontology Summit convention <uos-convene@xxxxxxxxxxxxxxxx>
From: "John A. Bateman" <bateman@xxxxxxxxxxxxx>
Date: Tue, 28 Feb 2006 10:05:17 +0100
Message-id: <4404124D.3060804@xxxxxxxxxxxxx>
Dear all,    (01)

sorry to have been delayed into wading into this discussion somewhat
late... never takes much to get a bunch of ontologists discussing! :-)    (02)

I will try and keep this as brief as possible (also for the above
reason and the correct sideswipe at the SUO-list that we have already
had onlist), but as a first statement it will tend to be a bit longer
than I like my emails to be; apologies for that.    (03)

Many of you (particular from the US) do not know the perspective
our group in Bremen takes on these issues so I will start with that,
drawing in a couple of positioning quotes from the discussion so far.
These can be followed on separate threads as relevant.    (04)

One word of motivation:    (05)

    our current work is coming from the area of spatial cognition,
    where we are researching/building systems that are spatially aware
    and can communicate/interact with humans during spatially-embedded
    tasks (http://www.sfbtr8.uni-bremen.de). A significant part of that
    is dealing with theories/models of space and spatial
    entities. Because of the connection with humans and getting
    interaction right, we have needed to take very seriously the entire
    area of qualitative spatial reasoning: there are many (10-20 at
    least) formalised/axiomatized qualitative spatial calculi that we
    take as starting points for our treatment of space.    (06)

First point of orientation:    (07)

    these treatments of space, which we take as contributing to the
    ontology of space, are NOT simply reducible to any one account. A
    reduction to say, Euclidean geometry, as the model throws out the
    baby with the bathwater: the point of the different calculi is that
    they support different kinds of inferences, with different kinds of
    computational complexity and different kinds of
    communication. (Simple example: region-based reasoning
    vs. path-based reasoning; one is good for knowing where you are in
    in terms of embedded and disjoint hierarchies of areas, the other
    for following a route and navigating.)    (08)

    First conclusion:    (09)

       We adopt as fundamental the notion of ontological
       perspectivalism as promoted by Barry. We do not seek necessarily
       to reduce to a common simple old-style ontology but prefer to
       work with (at least the formal possibility of) multiple
       ontologies represented as fully axiomatized theories. The extent
       to which some of theories that are related turn out to be
       trivially relatable (and hence placeable 'under' or treatable
       'as' a common ontology) depends on the theories and a lot of
       detailed work.    (010)

       We take the distinction between 3D and 4D ontologies to be of
       this kind and follow Barry in allowing this to take on a
       fundamental role. Our position is thus compatible with that
       apparently failed to be negotiated between Barry and Matthew
       West: two ontologies (of this kind) which are to be formally
       related and developed more or less independently as required (as
       Matthew suggests) but actually much real work will be done by a
       combination of both (as Barry suggests).    (011)

Second point of orientation:    (012)

     We need for our systems to deal not only with space, but with all
     the other features/properties/categories that are relevant for our
     spatially embedded agents. This brings us to the rest of the
     ontology outside of the space issue.    (013)

     Barry suggests a top-level breakdown of categories which is well
     known as he has presented it consistently over many years. We
     believe that these distinctions are also absolutely crucial. We
     also believe, however, that it is essential to spell out in detail
     the kinds and natures of the entities that these dependents
     presuppose so that we know what dependencies they entail. The
     closest and most detailed working out of such dependencies that we
     know of (and we know of a lot! :-) is that of the DOLCE ontology,
     which we also adopt in its entirety plus extensions that we can
     talk about separately.    (014)

     Second conclusion and current work:    (015)

       In cooperation with Nicola at LOA, we are currently engaged in
       the full-scale computational specification of the first order
       axiomatization of DOLCE plus specifications of the OWL-DL
       approximations to DOLCE .    (016)

Tools:    (017)

     We are using the Common Algebraic Specification Language (CASL)
     plus associated tools for development of formal theories and for
     specifying relations between formal theories. CASL is a fully
     documented de facto standard for software engineering (IFIP
     Working Group 1.3).    (018)

     http://www.cofi.info/CASL.html    (019)

http://www.informatik.uni-bremen.de/agbkb/forschung/formal_methods/CoFI/CASL/    (020)

    On top of this is the Heterogeneous Tool Set (HETS), which provides
    support for integrating specifications drawing on a variety of
    logics. It is undergoing extensive development at Bremen; some of
    the more recent extensions being worked on include an extension
    'downwards' to achieve a Description Logic sublanguage of first
    order CASL (CASL-DL), which we are using for certain ontology
    submodules, and an extension 'upwards' ('sideways'?) for further
    varieties of modal logics.    (021)

    A first use of CASL for ontologies was presented in the paper by
    Klaus Lüttich and Till Mossakowsi at FOIS-2004.    (022)

    The closest relation to CASL in the ontology community is probably
    Common Logic, with several interesting differences that can be
    discussed separately plus strong tool support.  Till Mossakowski is
    one of the prime movers of CASL and developer of HETS and I suggest
    that he be added to the discussion panel as representing
    Bremen. (http://www.informatik.uni-bremen.de/~till/)    (023)

Issues:    (024)

    As the 'downwards' and 'upwards' metaphors above indicate, all of
    the logics included in CASL are organized in a lattice. All of the
    specifications of theories produced within CASL are similarly
    placed both within this metastructure and as lattices of modules
    within logics. This gives an answer to the thread question from
    Mike Uschold:    (025)

    >> Years ago, John Sowa talked about a wonderful idea of a lattice
    >> of theories.  To date, I am unaware of any work by John or
    >> anyone else that explores this idea in any detail.    (026)

    This is already our day-to-day practice, which Till can provide all
    the formal details of.    (027)

    Third conclusion:    (028)

    James Schoening:
    >> re we going to discuss the technical feasibility of
    >> interrelating multiple upper ontologies?  To issue a communique,
    >> we'll need a good handle on whether we believe this is
    >> technically feasible and will enable semantic interoperability.    (029)

    Our response: yes this is technically feasible and actually it is
    essential for good ontology development. Its further use for
    semantic interoperability is a major research goal; currently we
    use it, for example, for relating different theories of space, for
    relating theories of space to theories of other entities, etc. All
    work very much in progress but we believe the direction is clear.    (030)

    Matthew West:
    >> we will have to move to a lattice of theories approach, as
    >> described by John Sowa. Here, basic theories are stated
    >> separately, and can be combined in a mix and match approach.    (031)

    Absolutely correct. HETS takes this a bit further in that we also
    mix and match logics.    (032)

    Which is a good point to throw in:    (033)

    >> (1) Is there a difference between a formally-expressed logical
    >> theory and an "ontology"?  (2) If you answered YES to (1), then
    >> what is that difference?    (034)

    We agree with (4):    (035)

    Leo: To me a logical ontology (as opposed to weaker models) is both
         A) a logical theory, i.e., a theory expressed in a logic, and
         B) a logical theory which purports to be something about the
         world ...    (036)

Final(ly) words:    (037)

   We also agree with Mike Uschold's warning:    (038)

   >> Too often, people focus first on the term, and then argue
   >> incessantly about what its definition should be. What they are
   >> really discussing many different 'things of interest', many of
   >> which are subtle variations on each other, and that will not end
   >> up in the ontology. Such arguing is typically a highly
   >> inefficient way to spend time.    (039)

   Absolutely. We have stories to tell about the relationship to
   language, another one of our central research areas, but that can be
   a separate thread. This confusion of discussions of terms and
   discussions of 'entities' must be *the* major source of time wasting
   in ontology discussions over the past 20 years.    (040)

That should show where we are coming from and that we are already in
agreement with many of the views that have been put forward; more
detailed info can go in smaller emails and on the WIKI and at the
meeting.    (041)

Best,
John B.
 _________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/uos-convene/
To Post: mailto:uos-convene@xxxxxxxxxxxxxxxx
Community Portal: http://ontolog.cim3.net/
Shared Files: http://ontolog.cim3.net/file/work/UpperOntologySummit/uos-convene/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UpperOntologySummit    (042)
<Prev in Thread] Current Thread [Next in Thread>