Dear Cory, (01)
My intended use in my language of the word "system" relates to "system
engineering" which, as I believe you also use the term (hope I am not
off track), is an abstraction not a technology, that determines what is
a system. In software, I agree on MDA/OMG etc as well as the agent
based societal approaches. (02)
Systems are structures
with unifying behavior. (03)
I agree with the statement because the structural elements act together
in way that is purposeful --- chaotic systems (like the chemcial
systems) are also purposeful if we extend the notions of purpose to the
rules governing equilibria (ie. purpose is to reah equilibrium or steady
state). Hence that suggests that all systems are subject to constraints
or equivalently, that they are parameterized by observables that are
either (i) visible or (ii) hidden. (04)
I am interested in identifying methods, starting from ontologies, to
identify (ii) by using ontologies as indicators of gaps in knowledge
(not as interested in what successful ontologies do say but more about
what they don't say and *why*). That is a purely research effort but I
wanted to indicate my motivations in that as developing methods (like
metaphors and anlogies) to help identify these gaps. (05)
PS - I am now using the dev list for this (I hope this is appropo) (06)
Thanks, (07)
Arun
Cory Casanave wrote: (08)
>What is a "system"?
>We should be careful in using such words as "system", in our enterprise
>architecture work we are very explicit that "system" is not limited to
>technology. An enterprise is a system, as is a business process, as are the
>computer programs that are a part of these larger systems. People,
>computers and things are all part of these systems. Systems are structures
>with unifying behavior.
>This is important because we DO NOT want to tie our terms of interaction to
>the technology; we want these to define the domain and the technology to
>take roles within that domain to provide value. There is established (MDA)
>technology for going from such domain models to information system models.
>Considering interactions at the computer level will only build yet another
>EAI solution.
>>From this view, interaction between systems is quite a general concept. It
>would cover, for example, how the US and UK work together for a joint
>mission.
>Can we describe systems?
>However, (shields up), it is not something that can be directly represented
>in FOL. We may simply need to say things that do not fit within this logic,
>and what we specify about systems should not be so constrained. It is
>modal, non-monotonic and probably a few other things that give FOL
>nightmares. This seems to be the advantage of the modular approach, we can
>have FOL definitions for statements that are valid in that logic, but allow
>others as well.
>As brain-dead as OWL+SWRL may be, you can write rules in SWRL that go beyond
>FOL. I don't know that CL can do everything OWL+SWRL can do either.
>Perhaps really smart people will eventually figure out the logic with the
>needed generality, but in the interim we need a way to have and manipulate
>well defined terms and concepts that go beyond the capabilities of the FOL
>approach.
>
>-----Original Message-----
>From: Arun Majumdar [mailto:arun@xxxxxxxxxxxx]
>Sent: Wednesday, January 18, 2006 12:29 AM
>To: Obrst, Leo J.
>Cc: Cassidy, Patrick J.; Smith, Barry; John F. Sowa; Adam Pease; Anne
>Cregan; Arsic, Antoinette; Brand Niemann; Charles Turnitsa; Chris Menzel;
>Christopher Spottiswoode; Cory Casanave; Dagobert Soergel; David Eddy; Eric
>Peterson; Gary Berg-Cross; Hans Teijgeler; Harold Frisch; I.N. Sarkar; J. P.
>Morgenthal; James Schoening; Jeffrey A. Schiffel; John Thompson; John Young;
>Ken Ewell; Lowell Vizenor; Matthew R. West; Michael Denny; Michael
>Gruninger; Nicolas Rouquette; Olivier Bodenreider; Pawel Garbacz; Peter Yim;
>Richard Lee; Richard Murphy; Roberto Bordogna; Roy Roebuck; Steve Ray; Walt
>Truszkowski; Wu Hanxin; John Cabral; Michael Gruninger
>Subject: Re: Before we start...
>
>Dear Leo,
>
>You wrote the following (to which I strongly agree and have some
>comments following)
>
>
>
>>I would use some variant of Common Logic (and possibly IKL from the
>>IKRIS project, once that is fully specified), because it (they) as
>>first-order logic languages are more expressive formally than >OWL+SWRL.
>>
>>
>
>In my current commercial work elsewhere, the most troublesome aspect
>that I face in interoperability projects is the difficulty defining a
>theory that directly supports a first-order manipulation of theories and
>models, independently from the particular foundational ontology that it
>is plugged in: Common Logic is an excellent first step, however, all
>ACL's for knowledge exchange require the latter (theory) to succeed
>especially if the effects of the interchange themselves change the
>behavior of the state of the system, which inevitably they will do in
>any investigative reasoning process that supercedes stove-piping and
>promotes evolvability, abduction, induction and adaptability etc...
>(agents or systems that cooperate, collaborate, communicate, coordinate,
>or affect command and control in the elements in a problem solving
>excercise).
>
>Any reactions appreciated.
>
>Thanks,
>
>Arun
>
>
>Obrst, Leo J. wrote:
>
>
>
>>Pat,
>>
>>I'll have more comments when I get a bit of time.
>>
>>Just on these recent comments:
>>For ontologies, Class is more accurate than Type, because we really
>>don't want to clash with type theory in computer science, which I think
>>we would. Type theory in the latter contain functional types, possibly
>>as higher-order as you wish to go.
>>
>>Also, the nature of the time-indexing of instances will vary depending
>>on whether you have a 3D or 4D ontology. For the latter, your
>>time-indexing could fall out as mereological relations between
>>"instances", all of which we would say have the same identity.
>>
>>In natural language semantics, index and context are usually
>>distinguished, with time, place, world, etc., being indices that
>>characterize a context. But many kinds of things can characterize a
>>context, i.e., security marking, provenance information, etc.
>>
>>Disjointness is probably better characterized as an integrity
>>constraint, but I would not necessarily exclude it in advance.
>>
>>I would use some variant of Common Logic (and possibly IKL from the
>>IKRIS project, once that is fully specified), because it (they) as
>>first-order logic languages are more expressive formally than OWL+SWRL.
>>
>>There have been many time theories proposed, both ontologies and
>>temporal logics. OWL-time, for example, and the set of theories that
>>Pat Hayes brought together a few years ago. I had a summary of this
>>back a couple of years on the SUO list: I'll try to dig it up.
>>
>>Thanks,
>>Leo
>>
>>
>>_____________________________________________
>>Dr. Leo Obrst The MITRE Corporation, Information Semantics
>>lobrst@xxxxxxxxx Center for Innovative Computing & Informatics
>>Voice: 703-983-6770 7515 Colshire Drive, M/S H305
>>Fax: 703-983-1379 McLean, VA 22102-7508, USA
>>
>>
>>-----Original Message-----
>>From: Cassidy, Patrick J.
>>Sent: Tuesday, January 17, 2006 3:32 PM
>>To: Smith, Barry; John F. Sowa
>>Cc: Adam Pease; Anne Cregan; Arsic, Antoinette; Arun Majumdar; Brand
>>Niemann; Charles Turnitsa; Chris Menzel; Christopher Spottiswoode; Cory
>>Casanave; Dagobert Soergel; David Eddy; Eric Peterson; Gary Berg-Cross;
>>Hans Teijgeler; Harold Frisch; I.N. Sarkar; J. P. Morgenthal; James
>>Schoening; Jeffrey A. Schiffel; John Thompson; John Young; Ken Ewell;
>>Obrst, Leo J.; Lowell Vizenor; Matthew R. West; Michael Denny; Michael
>>Gruninger; Nicolas Rouquette; Olivier Bodenreider; Pawel Garbacz; Peter
>>Yim; Richard Lee; Richard Murphy; Roberto Bordogna; Roy Roebuck; Steve
>>Ray; Walt Truszkowski; Wu Hanxin; John Cabral
>>Subject: RE: Before we start...
>>
>>At this point I just want to make a few comments, and may have more
>>time to elaborate this evening.
>>
>>As for "Type" in preference to "Class", I can accept that though I am
>>fond of "Class". Will the Cyc "Collection" be equivalent to "Type" in
>>this terminology?
>>
>>Except for the question of time-indexed "isa" (instance-of) I think
>>that I agree with everything else that has been said. More about time
>>indexing at (4) below.
>>
>>(1) For example about argument restrictions:
>>
>>
>>
>>
>>
>>>[AP] SUMO includes "domain", "domainSubclass", "range" and
>>>
>>>
>>>
>>>
>>"rangeSubclass"
>>
>>
>>
>>
>>>to support this.
>>>
>>>
>>>
>>>
>>>
>>[PC} and OpenCyc has argIsa and argGenls, and resultIsa and resultGenl,
>>which I believe are equivalent. I do prefer the specification of the
>>argument location within the predicate, e.g. arg1Isa, because it makes
>>the relation binary and for a consensus ontology I think that simpler
>>is better (i.e. fewer arguments may be easier to understand at a
>>glance). But of course it will be necessary to define a separate such
>>constraint predicate for each possible argument location. I can go
>>either way.
>>
>>(2) disjoint relation
>>[JC] 1. We should include disjointness assertions about the
>>types/kinds in the ontologies. These will be essential for merging
>>taxonomies since they will be a central source of inconsistencies.
>>
>>Yes, I agree, such axioms are important in specifying meaning for those
>>types that are not fully defined by N&S conditions. There is an
>>interesting question as to whether "disjoint" axioms should ever be
>>involved in inferencing other than at the time of data entry. Ontology
>>Works, for example, treats them as "integrity constraints" during
>>knowledge entry time, but they are not used during query time. But
>>this is an issue for implementation. Such predicates can be included
>>in the ontology, and if an implementation chooses to use them only as
>>"integrity constraints", they can be marked as such automatically.
>>
>>(3) [AP] A hierarchy view just provides an initial overview. I
>>think the OWL version will lead us astray in fact, because it leaves
>>out most of the content of a formal definition. It's suitable for
>>some kinds of interchange however, once systems have agreed on the
>>formal semantics of terms.
>>
>>[PC] I agree with Adam that OWL is woefully insufficient to accomplish
>>meaning specification at the level that is needed for semantic
>>interoperability. So I anticipate that a FOL version (I like SKIF --
>>any objections?) will have to be maintained as soon as any semantic
>>relations are added in. I will try to do that, if no one else
>>volunteers to help.
>>
>>But I also agree that an OWL version can be useful, even though it will
>>be lobotomized relative to the FOL version, because the existing tools
>>(Protege + Pellet) make experimentation with reasoning simple, and
>>because it will provide a link to a lot of work being done by others,
>>at least until some rule language becomes accepted in the Semantic Web
>>community. I will try to maintain that version if no one else
>>volunteers. I think that it will be useful if, as Adam puts it,
>>"systems have agreed on the formal semantics of terms". What I would
>>recommend is that anyone working with an OWL version be sure that there
>>is, for every term in the OWL version, some well-defined term in an
>>FOL version that is intended to mean exactly the same thing -- i.e. one
>>working with OWL "accepts" the more detailed semantics of the more
>>expressive version. The documentation - linguistic definitions -
>>should be identical for both versions, though parts of that intended
>>meaning will not be expressible in OWL.
>>
>>(4) time-indexed "isa"
>>[JC] 2. As Barry Smith pointed out, let's also address the basic
>>predicates
>>for something being an instance of a type/kind or for one type/kind
>>being
>>a more specific type of a more type/kind.
>> For example, cyc::isa is temporally qualifiable and that
>>information
>>is represented on the context of the sentence. How does that
>>correspond
>>to other ontologies "instance of" predicates?
>>
>>[AP] > Is Cyc:isa temporally qualified in a way other than the presence
>>of
>>
>>
>>
>>
>>>an instance statement in a temporally qualified microtheory? I'm not
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>>sure that temporally qualifying an instance statement is a good idea,
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>>although most other statements would benefit.
>>>
>>>
>>>
>>>
>>[JC] Temporal qualification is an independent feature that applies to
>>Cyc::isa.
>>So, isa sentences are not necessary temporally qualified, it's just
>>that
>>they can be qualified. (Note: cyc::isa probably didn't inspire
>>temporal
>>qualification but because there are a number of properties that can
>>easily be treated as collections and because those properties can
>>change
>>(i.e., your occupation, your education level, your marital status) then
>>
>>temporal qualification makes sense. There are other representational
>>strategies. They can probably be integrated. I'm just bringing this
>>up
>>because it is an important issue for defining the core.
>>
>>[PC] This is very important, and I think it should be discussed and
>>resolved immediately, as it is a core issue and does not depend on the
>>specifics of the hierarchy we may accept.
>>
>>One thing I am wondering about is whether the computational devices
>>used for dealing with time-indexed assertions are in fact any different
>>
>>
>>from the devices used for dealing with other types of "context". I
>
>
>>noticed that certain Cyc "microtheories" are in fact some
>>implementation of time intervals. I have also noticed that there is a
>>conference "TIME 2006", scheduled for June, dealing, among other
>>things, with temporal reasoning. Does Cyc or anyone else already have
>>what they consider and adequate way to handle time? Would it work with
>>other types of context (location? hypothetical situation?).
>>
>>Pat
>>
>>
>>Patrick Cassidy
>>MITRE Corporation
>>260 Industrial Way
>>Eatontown, NJ 07724
>>Mail Stop: MNJE
>>Phone: 732-578-6340
>>Cell: 908-565-4053
>>Fax: 732-578-6012
>>Email: pcassidy@xxxxxxxxx
>>
>>
>>-----Original Message-----
>>From: Smith, Barry [mailto:phismith@xxxxxxxxxxx]
>>Sent: Tuesday, January 17, 2006 12:56 PM
>>To: John F. Sowa
>>Cc: Adam Pease; Cassidy, Patrick J.; Anne Cregan; Arsic, Antoinette;
>>Arun Majumdar; Barry Smith; Brand Niemann; Charles Turnitsa; Chris
>>Menzel; Christopher Spottiswoode; Cory Casanave; Dagobert Soergel;
>>David Eddy; Eric Peterson; Gary Berg-Cross; Hans Teijgeler; Harold
>>Frisch; I.N. Sarkar; J. P. Morgenthal; James Schoening; Jeffrey A.
>>Schiffel; John Thompson; John Young; Ken Ewell; Obrst, Leo J.; Lowell
>>Vizenor; Matthew R. West; Michael Denny; Michael Gruninger; Nicolas
>>Rouquette; Olivier Bodenreider; Pawel Garbacz; Peter Yim; Richard Lee;
>>Richard Murphy; Roberto Bordogna; Roy Roebuck; Steve Ray; Walt
>>Truszkowski; Wu Hanxin; John Cabral
>>Subject: Re: Before we start...
>>
>>
>>
>>
>>
>>
>>>I find the word _class_ very confusing, since it has too many
>>>definitions. In some versions of set theory, it means something
>>>so big that it isn't a set (but none of those things are ever
>>>going to be implemented in our computer systems). In object
>>>oriented languages, it means something that has associated
>>>procedural attachments. And some logicians use the word _class_
>>>as a synonym for set. It's a hopeless mess.
>>>
>>>I strongly urge that we replace the word _class_ with something
>>>less confusing. I recommend the term _type_ for the intensional
>>>category and _set_ for the instances that belong to a type (or
>>>for just any random selection, which might not belong to a type).
>>>
>>>John Sowa
>>>
>>>
>>>
>>>
>>I strongly agree with John as to use of 'type' here.
>>BS
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>
>
>
>
> (09)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-dev/
To Post: mailto:ontac-dev@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config: http://colab.cim3.net/mailman/listinfo/ontac-dev/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (010)
|