[Top] [All Lists]

Re: [ontolog-forum] Context, at last!

To: Nicolas F Rouquette <nicolas.rouquette@xxxxxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Chris Menzel <cmenzel@xxxxxxxx>
Date: Mon, 20 Jun 2005 19:24:22 -0500
Message-id: <20050621002422.GB44275@xxxxxxxx>
On Mon, Jun 20, 2005 at 09:42:22AM -0700, Nicolas F Rouquette wrote:
> >True enough, so a neutrino's ontology of your front room will
> >probably look a lot different than your front room ontology!  Note,
> >though, that the notion of context is often just assumed in the
> >background of an ontology -- we often create ontologies *from the
> >perspective of a given perceiver or set of perceivers*, e.g., domain
> >experts in a mfg shop floor.  For these cases, the notion of context
> >needn't play an explicit role.  More and more, though, we have a need
> >to represent multiple ontologies arising out of different contexts
> >within a *single* framework, and this requires explicit mechanisms
> >that enable us to group pieces of information according to context --
> >and that's exactly what formal theories of context are attempting to
> >provide.
> ...
> What does it mean to use a *single* framework for representing
> multiple ontologies according to different contexts or some other
> criteria?     (01)

Roughly, a framework in which one can say such things as "According to
ontology 01, nuclear plant P has adequate security, but according to
ontology 02 P does have adequate security."  Or: "Given the information
in *both* 01 and 02 (but not either separately), the security for
nuclear plant P is adequate."      (02)

> What is the role of the framework in this case?    (03)

One role is simply to faciliate reasoning that is based on multiple
ontologies.  But a particularly important role for this sort of
framework concerns provenance, which is critical for assessing
confidence and hence for determining how we should act in response to
information from *this* or *that* particular source.    (04)

> Now, suppose we wanted a single framework with which we can represent
> multiple ontologies according to different contexts / perceivers.
> Streching Duane's examples a bit, we might have:
> - two ontologies about the same subject matter, i.e., the concept of a 
>   living room
> - neutrino ontology ( a bunch of empty space....)
> - couch potato ontology (somewhere to put a glass & a bowl of popcorn on 
>   top of)
> What does having a single framework mean here?    (05)

Same as in any other case.    (06)

> We don't have a lot of insight into people's relationships with
> neutrinos; therefore, the significance of a single framework for both
> ontologies isn't obvious nor intuitive.     (07)

So it's not a case where a single overarching framework would be
particularly useful.  That's simply a problem with the example, not the
idea. :-)    (08)

> Let's pick two ontologies with intuitive connections between them, e.g.:
> - two ontologies about the same subject matter, i.e., the concept of a 
>   living room
> - furniture store ontology (features - strength, texture, look, cleaning -, 
>   prices, building materials - wood, metal, cloth, ... -)
> - couch potato ontology (somewhere to put a glass & a bowl of popcorn on 
>   top of, stretch the feet on and kick the bowl in a moment of excitement)
> Ok, so spills can happen, someone's going to be upset if the table
> gets a stain from the drink or the popcorn that won't go away or if
> the heavy feet will break the table. These kinds of properties use
> both ontologies but in order to do so, it is necessary to establish
> correspondences between their concepts & individuals.     (09)

Right.    (010)

> What isn't clear to me is the following:
> a) defining a precise semantics for terms used to talk about ontologies of 
>    something ... (i.e., classes & object properties in OWL)    (011)

Well, that's just what a formal theory of contexts is supposed to
provide.  OWL currently does not provide any support for this -- nor was
it designed to in its current incarnation.    (012)

> b) using different logics for reasoning about and computing with the
> properties and property values of indidividuals and concepts in the
> ontology (i.e., data properties in OWL)    (013)

That's an important issue, for sure, but I take reasoning to be a
separate issue from representation.  At this point, by a formal theory
of contexts I'm talking about the problem of getting the
representational stuff figured out.  You need to know clearly *what* you
are reasoning about before you can expect to be able to reason
effectively about it.    (014)

> - logic of time (e.g., are all contexts defined w.r.t. the same clock or 
>   can we have 2 contexts defined w.r.t. 2 separate clocks)    (015)

Certainly a robust representation of time, in the calendar sense, is
essential -- we need to know when events occur, how long they take,
whether one occurs before another, etc.  Your reference to *clocks*
however, suggests relativistic phenomena, which (along with quantum
phenomena) by my dim lights don't have any role whatever to play in
99.99% of all ontologies, which are mostly about the affairs of slow
moving, medium-sized, dry goods. :-)    (016)

> - logic of space (e.g., something like newtonian physics - glass, popcorn, 
> table, ... - vs. something like quantum physics - neutrino & other exotic 
> particles -)    (017)

As noted!    (018)

Chris Menzel    (019)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (020)

<Prev in Thread] Current Thread [Next in Thread>