RE: SUO: RE: A proposed SUO content outline - roles
Pat, Ian,
You wrote:
> >
> > Functional Physical Object
> >
> > Contains axioms and relations that apply to all physical
> > objects that have functional, but need not have material
> > continuity, e.g. the Chairman of Shell.
>
>I'm somewhat confused by your example. It seems to me that the concept
>"Chairman of Shell" can be understood in two different ways, either as the
>class of all people who have assumed this role or as the person who
>currently holds this title.
I think neither of these is intended. The point is that the
chairmanship (like the US presidency, various categories of royalty,
etc.) is temporally continuous and always identified with a single
person at a time. It cannot be identified either with a person or
with a set. It is a distinct category with its own rules (as in 'the
king is dead: long live the king!') Restricting yourself to people
and sets is too limiting: there are more kinds of thing in the world
than just those two.
I think there is a third alternative. In a 4-D ontology these kinds of roles
correspond to physical objects (see my book Chap 7, Sect 2.3 for an extended
discussion) - where stages of people's lives correspond to stages in the
role's life (and no need for a new category of new objects). This was noted
as one of the benefits of Perdurantism in my note).
So it is more that there are two different kinds of physical object, than
two different kinds of object.
Regards,
Chris
-----Original Message-----
From: owner-standard-upper-ontology@ieee.org
[mailto:owner-standard-upper-ontology@ieee.org]On Behalf Of pat hayes
Sent: 27 February 2001 20:08
To: Ian Niles
Cc: standard-upper-ontology@ieee.org
Subject: Re: SUO: RE: A proposed SUO content outline
> >
> > State
> >
> > Contains axioms and relations that apply to all individuals
> > that have a non-zero temporal extent.
>
>I don't see how "State" differs from "Individual". Since all individuals
>exist in space/time, wouldn't they necessarily have a non-zero temporal
>extent?
No. Something can be temporally located but have a zero extent.
Timepoints are the obvious example, but more 'physical' examples
might be instantaneous events like lightning flashes, or beginning
and endings of longer events.
> >
> > Period of Time
> >
> > Contains axioms and relations that apply to all individuals
> > that cover all of space between two points in time.
>
>The corresponding axioms and relations in the merged ontology can be found
>in the section "Temporal Definitions/Axioms".
>
> >
> > Activity
> >
> > Contains axioms and relations that apply to all individuals
> > that are something happening to bring about change.
>
>As I've mentioned in other emails, the subject of change has not been
>systematically addressed in the merged ontology. I'm researching the
>possibility of incorporating something like the Situation Calculus, and I
>would welcome any suggestions along these lines.
The situation calculus is hopelessly limited for a general physical
ontology. A process ontology would be a more generally useful overall
framework. There are a number of quite elaborate process ontologies
in the public domain, any of which would be better than the sitcalc.
At a minimum, we need notions of process and subprocesss;
combinations of processes into larger processes including temporal
sequencing, conditions, parallel operations and ideas like a process
which 'spawns' other processes. We need to be able to reason both
forwards and backwards in time in a uniform framework, and we need to
be able to talk about constraints on processes (and classes of
processes) in terms of temporal and spatial constraints at the
boundaries, things and materials involved in the processes (either
being changed by them or required for them), and things conserved by
the process.
> > Physical Object
> >
> > Contains axioms and relations that apply to all individuals
> > that is a distribution of matter, energy, or both.
>
>This would correspond to 'Object' in the merged ontology, which is a direct
>subclass of 'Physical'.
I think that merged-ont'Object' would be too restrictive. For
example, 'the pope from 1400 to 1700' could be a (single)
PhysicalObject, but it would have to be a lot of distinct m-oObjects.
>The node of 'Object' figures into many selectional
>restrictions in the merged ontology, which are expressed with the
>'nth-domain' predicate.
>
> >
> > Materialised Physical Object
> >
> > Contains axioms and relations that apply to all physical
> > objects that have essential continuity of material or energy,
> > e.g. Car N199FOW.
>
>I'm not sure what you mean by "essential continuity of material or energy",
>so I guess I'm unclear about how this class is distinguished from the
>superclass.
>
> >
> > Functional Physical Object
> >
> > Contains axioms and relations that apply to all physical
> > objects that have functional, but need not have material
> > continuity, e.g. the Chairman of Shell.
>
>I'm somewhat confused by your example. It seems to me that the concept
>"Chairman of Shell" can be understood in two different ways, either as the
>class of all people who have assumed this role or as the person who
>currently holds this title.
I think neither of these is intended. The point is that the
chairmanship (like the US presidency, various categories of royalty,
etc.) is temporally continuous and always identified with a single
person at a time. It cannot be identified either with a person or
with a set. It is a distinct category with its own rules (as in 'the
king is dead: long live the king!') Restricting yourself to people
and sets is too limiting: there are more kinds of thing in the world
than just those two.
> If it denotes a class, then it doesn't have
>"material continuity" for the same reason that no class or set has material
>continuity - they're abstract objects. However, since the elements of this
>class are people, they do have material continuity. If, on the other hand,
>"Chairman of Shell" denotes a particular person, then it does have
"material
>continuity", contrary to what you say. I hope you can clear up my
confusion
>here.
>
> >
> > Stream
> >
> > Contains axioms and relations that apply to all physical
> > objects that is material or energy moving along a path, e.g.
> > a traffic stream flowing on the freeway, some liquid flowing
> > in a pipe.
>
>This is a definite gap in the merged ontology, and I would welcome any
>contributions towards filling it.
Well, I did discuss an ontology of liquids a few years ago at some
length, using a 4-d ontology and focussing on boundary conditions.
There is also a fairly detailed set of axioms in CYC.
> >
> > Temporal Boundary
> >
> > Contains axioms and relations that apply to all individuals
> > that have a zero temporal extent.
>
>It sounds like "Temporal Boundary" corresponds to the class of 'Abstract',
That seems clearly wrong. Abstract things don't have a temporal
location, but temporal boundaries do.
>another direct subclass of 'Entity'. Some axioms regarding 'Abstract' can
>be found in the section "General Axioms" of the merged ontology.
>
> >
> > Point in Time
> >
> > Contains axioms and relations that apply to all temporal
> > boundaries that are across all space, e.g. 3pm 20th February
> > 2001 UTC.
>
>The definitions for the functions corresponding to the concepts of year,
>month, day, hour, second, etc. are covered at the end of the section
>"Temporal Definitions/Axioms" of the merged ontology.
Yes, but is there a single class for them all? And is there a class
of all timepoints?
> >
> > Event
> >
> > Contains axioms and relations that apply to all temporal
> > boundaries that mark the beginning or end of some state.
>
>This corresponds to the concept of 'DiscreteProcess' in the merged
ontology,
I disagree. There is nothing in the notion of 'event' which requires
the event to be a discreteProcess. Continuously changing events may
also have beginning and ending states.
>which is a direct subclass of 'Process' and which has 'Act/Action' as an
>immediate subclass. The documentation for 'DiscreteProcess' is as follows:
>
>(documentation DiscreteProcess "In a discrete process, which is typical of
>computer programs or idealized approximations to physical processes,
changes
>
>occur in discrete steps called events, which are interleaved with periods
of
>
>inactivity called states.")
>
>Currently there are no axioms for 'DiscreteProcess' in the merged ontology.
>
> >
> > Collection
> >
> > Contains axioms and relations that apply to all things that
> > have members.
>
>These axioms and relations are covered in the new, provisional section "Set
>Theory" of the merged ontology. I say "provisional", because SUO
>participants haven't seen this stuff before, and it is subject to
>revision/rejection.
But not all collections are sets, as has been recently extensively
discussed.
> >
> > Class
> >
> > Contains axioms and relations that apply to all collections
> > to which we attach significance.
>
>In other words, by "Class" you mean a set that forms a natural kind?
Ian, what do YOU mean by 'natural kind'? Do you expect that there
will ever be a logical definiton of this concept?
> If so,
>then your concept of "Class" corresponds exactly with the concept of
'Class'
>in the merged ontology.
>
> >
> > Class of Individual
> >
> > Contains axioms and relations that apply to all classes that
> > have only individuals as members, e.g. pump, car.
? Wha? This seems to say that there is only one pump and only one
car, but I can't beleive that is what was intended. On the other hand
I have no idea what WAS intended. (What would be an example of a
class which wasn't a class of individuals?)
>One idea that strikes me is that we could make 'Class' a direct subclass of
>'Set'.
Unfortunately this is exactly the opposite usage from that used
throughout mathematics, where every set is a class but not vice
versa. (Using a different sense of 'class', however.)
> This would allow us to apply all of the set-theoretic
>definitions/axioms to classes, as well as sets. This isn't the way things
>are currently structured in the ontology (currently, 'Set' and 'Class' are
>disjoint siblings of the parent class 'SetOrClass'), but it may be a
>convenient way of reusing the set-theoretic apparatus in the context of
>intensional classes.
What do you mean by an 'intensional class' ? Sets are extensional by
definition.
> >
> > Quantifiable Property
> >
> > Contains axioms and relations that apply to all Class of
> > Individual that can be mapped to a number and unit of measure.
>
>This is covered in the section "Quantities and Units of Measure" in the
>merged ontology. The key concept in the section is the function
>'MeasureFn', which is defined as follows:
>
>(instance-of MeasureFn BinaryFunction)
>(nth-domain MeasureFn 1 RealNumber)
>(nth-domain MeasureFn 2 Unit-Of-Measure)
>(range MeasureFn Measure)
>(documentation MeasureFn "This function maps a real number and a unit of
>measure
>to that number of units.")
Not all measures can be multiplied by real numbers, so this will need
to be modified to be fully general. Also, what is a 'unit-of-measure'
? There seems to be something wrong with this definition. If the
unit-of-measure defines the 'units' involved, then we do not also
need to have a special measure function to tell us what we are
measuring. So for example if we multiply one INCH by three we get
three INCHES automatically. All we need is the unit and normal
multiplication.
Jerry Hobbs has a very nice general theory of measuring scales, by the way.
> >
> > Role
> >
> > Contains axioms and relations that apply to all classes that
> > indicate the role some thing plays in an activity or relation.
>
>What you are calling a "Role" is covered in the section "Definitions of
>Basic Binary Relations" in the merged ontology.
>
> >
> > Status
> >
> > Contains axioms and relations that apply to all classes that
> > are characteristics or qualities that is described by discrete,
> > unordered values.
>
>I agree that this is an important category, and it is also a gap in the
>current version of the merged ontology.
I have no clear idea what you guys are agreeing on. Can anyone give a
couple of examples of what this means?
>
> >
> > Organisational Level
> >
> > Contains axioms and relations that apply to all classes that
> > define the level of structure, e.g. atom, molecule, cell,
> > organism.
>
>I'm not sure exactly what you mean here. Do you envision having a
predicate
>that would relate a class with a quantitative or qualitative attribute
>specifying the level of granularity of the class? This might be a good
>idea, and there is nothing like this in the current version of the
ontology.
>Note that the folks at the Ontology Group at ITBM-CNR have developed an
>ontology called "Granularity" to cover just this sort of representational
>need.
Everyone agrees that there is a need for this, but not many people
know how to do it!
<snip>
> >
> > Whole-Part
> >
> > Contains axioms and relations that apply to all relations
> > that indicate that one individual is a part of another.
>
>This is covered in the "Mereotopological Definitions/Axioms" section of the
>merged ontology.
I think the following categories show that the notion of 'part' here
is meant to be interpreted more broadly than in mereotopology.
> >
> > Fusion Whole-Part
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that the whole is precisely the sum
> > of the parts.
> >
> > Arrangement Whole-Part
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that the parts are arranged, but not
> > necessarily touching, in order to a create a whole that is
> > more than the sum of the parts.
> >
> > Assembly Whole-Part
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that the parts are assembled in a way
> > that leaves them touching to make a whole. The nature of an
> > assembly relation is that assembly and disassembly is simple
> > and non-destructive. E.g. the assembly of a spark plug as
> > part of an engine.
> >
> > Feature Whole-Part
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that the part is contiguous with the
> > whole. Feature whole-part is not necessarily separable, and
> > if separated, it cannot necessarily be easily reassembled.
> >
> > Temporal Whole-Part
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that the part is all of the spatial
> > dimensions of the whole, but only a temporal part of it.
>
>
>We currently have concepts in the merged ontology to cover the various ways
>in which an object may be composed out of its parts, e.g.
>'ContinuousObject', 'CorpuscularObject', 'OrganicObject', and 'Assembly'.
They are pretty poor, frankly. None of these have clear definitions
in the merged ontology, and clearly we are going to need more than
this.
>These are all high-level notions that are defined in the "General Classes"
>section of the merged ontology.
>
> >
> > Participation in Activity
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that a state of a whole individual
> > is a participant in an activity.
>
>I'm not sure how this category differs from your earlier category of
"Role".
>
> >
> > Temporal Bounding of State
> >
> > Contains axioms and relations that apply to all whole-part
> > relations that indicate that a temporal boundary bounds a
> > state.
>
>Wouldn't this class just consist of two functions, one from a state to the
>point in time at which it comes into being and one from a state to the
point
>in time at which it ceases to exist. If so, I can formally specify these
>functions and add them to the merged ontology.
The temporal boundary in 4d is not (usually) a point a time, but a
piece of a spatiotemporal surface.
>
> >
> > Containment of Individual
> >
> > Contains axioms and relations that apply to all whole-part
> > relations where the part is contained by the whole.
>
>Well, every mereological relation involves a notion of containment, in some
>sense of containment. How does this category differ from the earlier
>"Part-Whole" categories?
Imagine a plastic cup contining coffee. The cup contains the coffee,
but every mereotopological part of the cup is entirely full of
plastic, not coffee.
>
> >
> > Connection
> >
> > Contains axioms and relations that apply to all relations that
> > indicate that one individual is connected to another
> > individual.
>
>In the merged ontology we have the predicate 'connected' and the class of
>'SelfConnectedObject'. Both of these are covered in the "Mereotopological
>Definitions/Axioms" section of the merged ontology.
Consider a metal chain. The links are connected (in one sense) to
each other, but none of the links are connected to other links in the
mereotopological sense (sharing a subpart).
> >
> > Temporal Sequence
> >
> > Contains axioms and relations that apply to all relations that
> > indicate that one individual (i.e. its whole spatio-temporal
> > extent) precedes another.
>
>We do have a calculus of time intervals and their relations in the
"Temporal
>Definitions/Axioms" section of the merged ontology, and I think we could
>reuse this calculus to talk about the temporal relations between
individuals
>if we defined a function that maps individuals to the time interval during
>which they exist, e.g. as follows:
>
>(instance-of TimeIntervalAbstractionFn UnaryFunction)
>(nth-domain TimeIntervalAbstractionFn 1 Object)
(range TimeIntervalAbstractionFn TimeInterval)
I'd suggest a less barabaric name might be 'when' (and a similar
function from things to their locations called 'where')
> >
> > Involvement in Activity
> >
> > Contains axioms and relations that apply to all relations that
> > indicate that a thing is involved in an activity. The
> > difference between involvement and participation is that the
> > involved object is a class or not contemporaneous with the
> > activity, e.g. a discussion today about World War II.
>
>I wonder if all cases of what you are calling "Involvement" could be
covered
>by two predicates, viz. one linking a ContentBearingObject with a class
that
>describes a topic of the object and one linking a ContentBearingObject with
>a particular object that is a referent of the object.
>
> >
> > Set
> >
> > Contains axioms and relations that apply to all mathematically
> > defined sets.
The idea of a 'mathematically defined' set is a bad idea. There is no
clear category of a mathematical definition (as opposed to a
nonmathematical one), and in any case what is in a set isn't
determined by the form of the definition.
>This is covered in the new and provisional "Set Theory" section of the
>merged ontology.
>
> >
> > o-------------------------------------------------------------
> > --------------
> > ---o
> >
> > Regards
> > Matthew
> > ============================================
> > Matthew West
> > Operations & Asset Management
> > Shell Services International
> > H3229, Shell Centre, London, SE1 7NA, UK.
> > Tel: +44 207 934 4490 Fax: 7929
> > Mobile: +44 7796 336538
> > E-mail: Matthew.R.West@is.shell.com
> > http://www.shellservices.com/
> > ============================================
> >
>
>
>Attachment converted: Betelgeuse:Merge.txt 2 (TEXT/MSWD) (0002EF4A)
---------------------------------------------------------------------
IHMC (850)434 8903 home
40 South Alcaniz St. (850)202 4416 office
Pensacola, FL 32501 (850)202 4440 fax
phayes@ai.uwf.edu
http://www.coginst.uwf.edu/~phayes