SUO: RE: A proposed SUO content outline
Hi John,
I agree with you that one of the biggest challenges in constructing
an overarching ontology is that many of the various chunks that have to be
merged to create the ontology will be, to a lesser or greater extent,
incompatible. In some cases, the incompatibilities can be smoothed over by
tweaking the formalizations; in other cases, wholesale theoretical revision
may be required. Although we agree that this is a major challenge for our
project, I guess we disagree about how it should be tackled. From what you
say below, it sounds like you are recommending that we put a halt to the
actual ontology construction, take a step back, and get straight about the
philosophical foundations before we go any further. Well, this approach,
one might regard it as the top-down approach to ontology design, has been
tried, and, as you know, it comprises the metaphysical component of the
2,500 year history of western philosophy. Unfortunately, this approach has
not netted us any real conclusions. It has given us a nice KR language, the
first-order predicate calculus, but it has not yielded any content which a
majority of philosophers could agree on.
In any case, I would argue for a "reflective equilibrium" approach
to ontology design. Let's start with a set of high-level categories and
then successively incorporate new content into the ontology. As we discover
incompatibilities between new content and existing content, we can resolve
them on a case-by-case basis. If we find that some high-level categories
are not needed to support lower-level nodes, then we can excise them from
the conceptual structure. If we find that a lower-level structure can be
simplified by locating part of its semantic endowment in a new higher-level
node (from which the lower-level structure can inherit this content), we can
add that node to the ontology. It seems to me that this sort of approach
has the advantage of making the various incompatibilities between formal
theories as concrete as possible (as Wittgenstein once pointed out, many
philosophical muddles can be traced to a dearth of real examples).
Furthermore, it has the benefit of incrementally building up an engineering
artifact that could potentially be very useful for integrating information
systems, even if we have to neglect some philosophical niceties.
-Ian
> -----Original Message-----
> From: John F. Sowa [mailto:sowa@bestweb.net]
> Sent: Saturday, February 24, 2001 9:19 AM
> To: West, Matthew MR SSI-GREA-UK; Standard-Upper-Ontology (E-mail)
> Subject: Re: SUO: RE: RE: A proposed SUO content outline
>
>
>
> Dear Matthew and Ian,
>
> Every term on your outline is necessary for a complete system
> of knowledge representation. Indeed, all of these terms have
> been used for many years in mathematics and physics, and they
> have been formalized, axiomatized, and reasoned and computed
> with in great depth in many different ways for a long time.
>
> The big question is not whether they belong in the SUO. The
> answer to that is simple: yes, of course.
>
> Even the issue of writing definitions and axioms for them is
> not a big problem. There are plenty of axiomatizations for all
> of them that can be taken out of the literature of math, logic,
> and physics, dusted off, and translated into KIF, CGs, and
> many other notations.
>
> One issue that has to be recognized: all those axioms
> were designed for different purposes, and they use very
> different primitives. As an example, I use the Eulerian and
> Langrangian systems of coordinates for fluid mechanics. Many
> of the same phenomena can be represented in either one, but
> the representations have very different structures and
> relationships. That is true of all those categories in
> the outline: there are many, many different coordinate
> systems and representations that have been used, all of them
> are valuable for different purposes, but the relationships
> between them are very far from clear. Taking one coordinate
> system instead of another is a purely arbitrary choice, and
> you have to know how they are related before you can make
> a rational decision about how they should be positioned
> in the ontology.
>
> The first thing that has to be determined is which of these
> issues belong to the logic, which of them belong to the
> ontology, how the logic relates to the ontology, and how
> the logic and the ontology relate to the real world and/or
> any possible world, situation, state of affairs, etc. And
> the next thing is how can you reason about them with the
> logic and talk about them with language(s) you want to use.
>
> We all know that the knowledge representation must accommodate
> individuals, states, events, space, time, space-time, process,
> activity, agents, etc. But you cannot start to put these things
> into a hierarchy until you have a very clear idea of what these
> things are, how you can refer to them in logic, how they are
> classifiable by the ontology (Note: classifiable is a
> prerequisite to being classified), and how they relate to
> what is in the world, what is perceived by humans, animals,
> or robots, how evidence for them is obtained, etc.
>
> Only after these questions are answered, does it become
> possible to say where you can position the categories in
> the hierarchy, outline, or whatever you want to call it.
>
> Compared to that, questions like the following are a tiny,
> insignificant little nit:
>
> >MW: Do you have some things in the merged ontology that are
> not found here?
>
> >(e.g. holes).
>
> We still haven't answered where the concepts of Circle, Sphere,
> and Cube belong. When we know where those belong, then maybe
> we can begin to talk about holes in a cube or sphere.
>
> All of Ian's questions about the following list are well taken,
> but that is barely scratching the surface of the ones that must
> be addressed before we can put these topics into some ordering:
>
> >> > 1. Thing
> >> > 1.1. Individual
> >> > 1.1.1. State
> >> > 1.1.1.1. Period of Time
> >> > 1.1.1.2. Activity
> >> > 1.1.1.3. Physical Object
> >> > 1.1.1.3.1. Materialised Physical Object
> >> > 1.1.1.3.2. Functional Physical Object
> >> > 1.1.1.3.3. Stream
> >> > 1.1.2. Temporal Boundary
> >> > 1.1.2.1. Point in Time
> >> > 1.1.2.2. Event
> >> > 1.2. Collection
> >> > 1.2.1. Class
> >> > 1.2.1.1. Class of Individual
> >> > 1.2.1.1.1. Quantifiable Property
> >> > 1.2.1.1.2. Role
> >> > 1.2.1.1.3. Status
> >> > 1.2.1.1.4. Organisational Level
> >> > 1.2.1.1.4.1. Shape
> >> > 1.2.1.1.5. Information Pattern
> >> > 1.2.1.2. Class of Class
> >> > 1.2.1.2.1. Number
> >> > 1.2.1.2.2. Class of Relation
> >> > 1.2.1.2.2.1. Specialisation
> >> > 1.2.1.2.2.2. Unit of Measure Mapping of Property to Number Space
> >> > 1.2.1.2.2.3. Class of Representation
> >> > 1.2.1.2.2.3.1. Identification
> >> > 1.2.1.2.2.3.2. Definition
> >> > 1.2.1.2.2.3.3. Description
> >> > 1.2.1.3. Relation
> >> > 1.2.1.3.1. Classification
> >> > 1.2.1.3.2. Cause and Effect
> >> > 1.2.1.3.3. Whole-Part
> >> > 1.2.1.3.3.1. Fusion Whole-Part
> >> > 1.2.1.3.3.2. Arrangement Whole-Part
> >> > 1.2.1.3.3.3. Assembly Whole-Part
> >> > 1.2.1.3.3.4. Feature Whole-Part
> >> > 1.2.1.3.3.5. Temporal Whole-Part
> >> > 1.2.1.3.3.6. Participation in Activity
> >> > 1.2.1.3.3.7. Temporal Bounding of State
> >> > 1.2.1.3.3.8. Containment of Individual
> >> > 1.2.1.3.4. Connection
> >> > 1.2.1.3.5. Temporal Sequence
> >> > 1.2.1.3.6. Involvement in Activity
> >> > 1.2.2. Set
>
> Every level of this outline raises very serious questions about
> why things are placed on that level, how they are related to
> their siblings or their parents and children. Why, for example,
> is "period of time" under "state"? What is time? Why is
> "cause and effect" a sibling node of "whole-part"? What is
> causality? How does causality relate to time? Which is more
> fundamental, time or causality? Which one presupposes the
> other in its definition? Why?
>
> Having a laundry list of items to be considered is useful.
> But without a very serious answer to all these questions,
> it should not considered as anything that has any more
> structure than a laundry list.
>
> John Sowa
>
>