ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Ontology similarity and accurate communication

To: "[ontolog-forum] " <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Wed, 12 Mar 2008 12:33:13 -0500
Message-id: <p06230906c3fdb7a26e11@[10.100.0.20]>
At 11:45 AM -0400 3/12/08, Patrick Cassidy wrote:
Matthew,
  There are some points I think are worth pursuing:
[PC] > >  We know how to resolve terminology
> > clashes - use different names or different namespaces for the
> > different concepts.
>
> MW: That is not true for 3D/4D, it is about what sorts of things
> physical objects (in particular) are. For a 4-dimensionalist they
> are extended in time and have temporal parts, for a 3 Dimensionalist
> they wholly exist now, pass through time, and do not have temporal
> parts. The terminology is frankly irrelevant, it is the ideas that
> are diametrically opposed.

   As I have mentioned before, I believe that 3D and 4D are different and
compatible views of the same entities.

Here is the heart of the issue, I believe. Yes, in a sense of course this is correct. There are things in the actual world which everyone agrees we are all are talking about: people, events, timeperiods and so on. Nobody is disagreeing about that. So if these actual, real things are your 'entities' (as they probably are for the editors of Longmans) then we are all talking about the same things. However, to say this is not to say that we all conceptualize this one reality in the same way, or still less that only one conceptualization is possible, or that we will all come to agree on one. In fact, it is close to provable that more than one conceptualization is possible. (I say 'close to' because the possibility remains that if enough work were done, the alternatives might turn out to be isomorphic in a strong enough sense, just mapped to philosophical words differently: but I doubt this, myself, as deeper analysis never seems to converge to conceptual agreement.)

 We can use different words to
describe our views, but the important question for the machines is whether
the different views lead to *representations* that are logically
inconsistent.

Well yes, they do. And in fact the representations seem to come first: the words have been invented as terms of art in order to explain the conceptualizations.

 We don't need to agree on which aspect of the entity is of
greatest importance to us

Everyone is agreed on that point.
- we have to agree on how to represent those
different views, optimally in a way that can be related and not be logically
inconsistent.  I presented an example in a previous post in this thread of
how one 3D view can be reconciled with one 4D view.  If you think there are
4d and 3D views that are logically inconsistent, please show us the axioms
that lead to inconsistency.
   To say
      (1) extended in time and have temporal parts or
      (2) wholly exist now, pass through time, and do not have temporal
parts
  Is to describe two different views of the same object.

Yes, in a sense: but its those 'views' that get described by the ontologies, not the objects themselves. And if two people have divergent such views, and each think the other's view incoherent, neither of them is going to be happy with a third view which manages to reconcile them. Part of the basic assumption of each ontology are axioms which explicitly deny (or which make syntactically illegal) basic assumptions of the other.

 The task of
ontological engineering is not to describe the ultimate structure of
reality, but to create a representation that is useful for automated
inferencing in solving practical problems.  The only thing relevant is how
these views are formalized.  It can be helpful for people using the ontology
to have good documentation as well - the documentation can include the
different interpretations.  The issue is whether the formalizations are
inconsistent.

[[[3]]]
[MW] > MW: Yes, you could put both 3D and 4D theories into some ontology
> repository, and provide a mapping between them. This is quite possible.
> But make no mistake, you now already have a lattice of theories and
> two foundation ontologies, not one. Each physical object would be
> represented twice, once as a 3D object and once as a 4D object

  No, you would not have two ontologies, you would have one self-consistent
ontology in which two different and compatible views are represented, with
axioms that automatically convert an assertion in one form into the
equivalent assertion in the other form.

Well, that's a point of view, and I applaud it myself: but to phrase Matthew's point slightly differently: make no mistake, some of the users of your ontology will vehemently object to some of the inferences it makes, regarding them as philosophical errors or worse, and your ontology, even if it is self-consistent internally, will not be consistent with other ontologies that these folk prefer to use. So you will have to do the inter-ontology mappings that you are trying to avoid in any case, in order to achieve interoperability across communities.

 One could, of course, create
subontologies in which either of the forms is represented without the other.
But these will be logically compatible, and any time assertions in the two
subontologies need to be related, they can be converted accurately.

Actually it goes only in one direction, as the 4-d framework is more expressive than the continuant/occurrent framework, and can 'say' things that are incoherent or impossible in the other.

 In the
full ontology, one might very well have two different representations of the
same entity, but they would be created automatically and accurately from a
single assertion in either form (4D or 3D), and would not change the effects
of the inferencing.

Sure they would. They increase the size of the search space. Read Barry's posts to the ontolog thread on 'orthogonality'.

 Fear of redundancy is one of the factors that has
inhibited creation of such an inclusive ontology.  I do not believe it is a
problem, because utilities can be created to allow users to select only the
parts they need from the full ontology, and in that way the redundancy will
have no effect on performance in stand-alone applications.

Again, I applaud this idea, but the technical issues involved are essentially the same as those involved in describing mappings between ontologies. Its the same problem in a different terminology. And having all this in one large ontology puts extra strain on the logical language: it has a lot more to express. (This is why Cyc adopted 'micro-theories' to try to keep track of all the relationships between sub-ontologies. Terms change their meanings in different ontologies, for example.)

 When the full
ontology is used to enable interoperability between two views, the
inferencing may be slower.  How much slower will be discovered in practice.


[[[4]]] [MW] >> MW: Just take a close look at what Pat H is saying about the
> incompatibility between 3D and 4D.
>
I did, and I was astonished to see him produce a set of assertions in which
the '=' sign is used (it appears to me) in two different senses

OK, if that example bothers you (and I agree it is somewhat tendentious), try this, which makes the same basic point but more realistically, and is therefore more complicated.

First, a fragment of a continuant/occurrent ontology, such as DOLCE. Here, the categories of Occurrent and Continuant (think process and object respectively) are firmly and without exception asserted to be disjoint. One can speak of temporal parts of an occurrent by using during:
(during O t) is the temporal part of the occurrent O at the time t. Occurrents however cannot have temporal parts, so one would use fluent language to speak of a changing property of a continuant:
(P c t) rather than (P (during c t)). It would be natural to have a domain axiom for during:

(forall (x)(if (exist (y (t Time))(= y (during x t))) (Occurrent x) ))

Ie if something has a temporal part, then its an Occurent, from which it follows in this ontology that it is not a Continuant. {Note: this axiom as written would not do the job in CLIF, in fact, as all functions there are total. See the IKL Guide document for more details on how to do this properly, which I omit here for the sake of simplicity.} And certainly in this ontology, continuants exist: they are a central category, so for example we might have

(Continuant PatHayes)

OK, now turn to a '4-d' ontology. Here, all spatiotemporal entities have temporal parts, and the two forms of _expression_  (P x t)  and (P (x during t)) are completely equivalent, mere syntactic alternatives:

(forall (x (t Time) P)(iff (P x t)(P (x during t)) ))

Now, how do we put these together? Since the latter is simply more permissive than the former, we might just try to combine them directly, but then a contradiction arises whenever the second uses during on a continuant argument. Or, we could treat the continuant/occurrent distinction as being real even in the second ontology, rendering it consistent at the cost (unacceptable to its users) of making it effectively the same as the first one. Or, we could weaken the first ontology slightly, by removing the assumption of disjointness, making it effectively similar to the second one: but its devotees will object that this change utterly destroys the very distinction that they are at such pains to preserve, because it is so fundamental. There is no way to make everyone happy. Or, one can divide the universe into two sub-universes, one containing 4-d 'things' and the other containing the continuants and occurrents, restrict each sub-ontology to its part of this enlarged universe, and proceed: but now the two sub-ontologies are effectively isolated from one another, and the whole construct is a single ontology only in name, not in any useful sense. Ther are now two PatHayeses, the continuant and the 4-d one, and no way in the ontology itself to even state what the relationship might be between them.

, but is
considered identical for the purpose of inferencing.  He has denied that the
meanings are different

In passing: the equality symbol in CL is part of the logic, so obviously has the same meaning everywhere.

, and we will soon I expect proceed to discuss this
point further.

I look forward to it.

Pat
 But if that is the best argument against 3D being compatible
with 4D, I think my case is proven.

Pat

Patrick Cassidy
MICRA, Inc.
908-561-3416
cell: 908-565-4053
cassidy@xxxxxxxxx


> -----Original Message-----
> From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
> bounces@xxxxxxxxxxxxxxxx] On Behalf Of matthew.west@xxxxxxxxx
> Sent: Wednesday, March 12, 2008 10:20 AM
> To: ontolog-forum@xxxxxxxxxxxxxxxx
> Subject: Re: [ontolog-forum] Ontology similarity and accurate
> communication
>
> Dear Pat,
>
> >
> > MW - one request for clarification:
> >
> > [MW] >
> > > MW: I'm afraid I do not agree with Pat C at all. I have
> > spent much of
> > > my life doing business analaysis for systems design, and reviewing
> > > data models produced by others. I am always surprised at just how
> > > different (and usually limited) peoples mental models are.
> > I have yet
> > > to see two data models of the same application look the
> > same except by
> > > cut and paste.
> > >
> >
> > I am well aware that models created in isolation from each
> > other will vary
> > widely.  That is a restatement of the problem.  The hypothesis of the
> > 'Conceptual Defining Vocabulary' states that it will be
> > possible to solve
> > that problem with a common ontology of agreed basic concepts
> > that are used
> > to specify the meanings of the more specialized concepts in
> > the different
> > models.
>
> MW: I agree it is possible to come up with a common ontology of
> basic concepts that can be used to integrate the diverse models
> you talk about. I have done this myself with ISO 15926. However,
> you are claiming that there is ONE such ontology which we all
> share by some sort of osmosis. The point I was trying to make
> about all the different data models is that there is not such
> ONE ontology. I beleive there are several possible ontologies with
> different ontological foundations, each of which would be quite
> capable of helping to integrate diverse other ontologies of
> whatever form.
>
> MW: Further you are claiming that there are a limited set of
> foundation objects from which all others can be defined. This
> also does not sit true with me. I have seen new primitive
> concepts arise at almost any and every level of an ontology. I
> agree that it is very useful when classes can be defined as the
> intersection of some other classes, but it doesn't actually happen
> that often for me (though I admitedly usually leave these to
> be implied, rather than making them explicit).
>
> MW: Just as a simple example, can you show how you would move
> from even a relatively general class like pump, to centrifugal
> pump? Where do you get the centrifugal from?
> >
> > **  The relevant question is, when a group of people who are
> > determined to
> > find a common model of *basic* concepts get together, what
> > kinds of issues
> > remain that cannot be resolved by sincere efforts to find ways to
> > accommodate the needs of all of them? **
>
> MW: My experience again is that whatever you choose, there will
> be some people who will find it unpalatable enough to walk away.
> I would certainly not waste my time with anything 3D as an example,
> and there are others who would not countenance 4D.
> >
> >   I am very interested in finding out just what kinds of
> > residual problems
> > there really are.  Thus far the examples I have seen all resolve to a
> > terminology clash - two different people want to use the same
> > term to refer
> > to concepts of different meaning.  We know how to resolve terminology
> > clashes - use different names or different namespaces for the
> > different
> > concepts.
>
> MW: That is not true for 3D/4D, it is about what sorts of things
> physical objects (in particular) are. For a 4-dimensionalist they
> are extended in time and have temporal parts, for a 3 Dimensionalist
> they wholly exist now, pass through time, and do not have temporal
> parts. The terminology is frankly irrelevant, it is the ideas that
> are diametrically opposed.
>
> > If it is important to relate those concepts to
> > each other, that
> > takes a bit of work to analyze the reasons for the
> > differences and find the
> > relations between the two different representations.
>
> MW: Yes, you could put both 3D and 4D theories into some ontology
> repository, and provide a mapping between them. This is quite possible.
> But make no mistake, you now already have a lattice of theories and
> two foundation ontologies, not one. Each physical object would be
> represented twice, once as a 3D object and once as a 4D object.
> >
> >   If there is a case where an effort of that kind could not arrive at
> > agreement on how to include or reconcile some seemingly logically
> > incompatible representations, I would very much like to learn
> > what those
> > problems are, in detail.
>
> MW: Just take a close look at what Pat H is saying about the
> incompatibility between 3D and 4D.
> >
> > Pat
> >
> > Patrick Cassidy
> > MICRA, Inc.
> > 908-561-3416
> > cell: 908-565-4053
> > cassidy@xxxxxxxxx
> >
> >
> > > -----Original Message-----
> > > From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
> > > bounces@xxxxxxxxxxxxxxxx] On Behalf Of matthew.west@xxxxxxxxx
> > > Sent: Wednesday, March 12, 2008 6:41 AM
> > > To: ontolog-forum@xxxxxxxxxxxxxxxx
> > > Subject: Re: [ontolog-forum] Ontology similarity and accurate
> > > communication
> > >
> > > Dear John and Pat H.,
> > >
> > > > >In other words, all the axioms are at the task level, and each
> > > > >message sent between systems identifies what ontology is
> assumed.
> > > >
> > > > I agree that makes a certain sense, but Im less sanguine
> > than you are
> > > > about being able to neatly express relationships between
> > ontologies.
> > > > Not that such relations are impossible, but I think they
> > will will be
> > > > messier and more tangled, as Mala says in her recent
> > message. Which
> > > > is not necessarily a problem or something to avoid, just
> > something we
> > > > should be ready for.
> > >
> > > MW: I agree with Pat here, that just encouraging an
> > unlimited number of
> > > ontologies and saying we will map between them, is perhaps
> > where we are
> > > headed at the moment, but it is an expensive direction to
> > take. My best
> > > hope at the moment is to encourage convergence onto a limited
> number
> > > of ontologies - say 10 or so. They would have clearly stated
> > > foundations
> > > where the differences would be known and understood, and
> > mappings could
> > > be provided. I think several is good, because that provides
> > a market,
> > > which will help to drive improvement. It might also be that
> > some die,
> > > and others are born.
> > >
> > > MW: I'm afraid I do not agree with Pat C at all. I have
> > spent much of
> > > my life doing business analaysis for systems design, and reviewing
> > > data models produced by others. I am always surprised at just how
> > > different (and usually limited) peoples mental models are.
> > I have yet
> > > to see two data models of the same application look the
> > same except by
> > > cut and paste.
> > >
> > > Regards
> > >
> > > Matthew West
> > > Reference Data Architecture and Standards Manager
> > > Shell International Petroleum Company Limited
> > > Registered in England and Wales
> > > Registered number: 621148
> > > Registered office: Shell Centre, London SE1 7NA, United Kingdom
> > >
> > > Tel: +44 20 7934 4490 Mobile: +44 7796 336538
> > > Email: matthew.west@xxxxxxxxx
> > > http://www.shell.com
> > > http://www.matthew-west.org.uk/
> > >
> > >
> > > _________________________________________________________________
> > > Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> > > Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-
> > > forum/
> > > Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> > > Shared Files: http://ontolog.cim3.net/file/
> > > Community Wiki: http://ontolog.cim3.net/wiki/
> > > To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
> > >
> >
> >
> > _________________________________________________________________
> > Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> > Subscribe/Config:
> http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>
>
>
>
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-
> forum/
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/
> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>

 
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/ 
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/ 
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
 


-- 
---------------------------------------------------------------------
IHMC               (850)434 8903 or (650)494 3973   home
40 South Alcaniz St.       (850)202 4416   office
Pensacola                 (850)202 4440   fax
FL 32502                     (850)291 0667    cell
http://www.ihmc.us/users/phayes      phayesAT-SIGNihmc.us
http://www.flickr.com/pathayes/collections


_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (01)

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