Condensed SUO Correspondence Through 13 June, 2001

wattlexanaduΛογισμικό & κατασκευή λογ/κού

31 Οκτ 2013 (πριν από 3 χρόνια και 10 μήνες)

573 εμφανίσεις

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
1












Condensed SUO Correspondence

Through 13 June, 2001

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
2


1

Introduction & Miscellaneous

5

1.1

Terminology

8

1.1.1

Ontology

10

1.1.2

Context

10

1.1.3

Metonymy

11

1.1.4

Semiot ics, Zoosemiotics, Message

12

1.2

Lexical Items/Terms/…, Polysemy, …

13

1.3

Ontological Entities vs. Lexic
al Entities and Triadicity

14

1.4

2001, March 6
--

The Story So Far

15

1.5

2001, May 22


Now what?

16

2

Ontology: The State of the Art

17

3

Ontology Development Methodol ogy

18

4

Ontological Architectural Concerns

33

4.1

The [Tip
-
]Top Level

35

4.2

Mereology versus Set Theory

3
9

4.3

Collection, Aggregation, Set, Class, Mereosum, Fusion, Group, …

39

4.4

Domains of Interest

40

4.4.1

Point of View; World View

43

4.4.2

Peaceful Coexistence

of Consequences

44

4.4.3

Dividing and Conquering Logic

49

4.5

Contexts

50

4.6

Merging

50

4.7

Situations

51

4.8

Possible Worlds, Existence as a Predicate, Possibilia' vs. Bubbles, ‘IsReal’ & Otherwise

51

4.9

Conceptual Parsimony versus Possible Worlds

57

4.10

Kent’s Information Flow Framework (IFF) an
d Truth Concept Lattice

59

5

Outsi de
-
the
-
Box Noti ons of Scope and Purpose

60

5.1

Research

60

5.2

Desirability/Possibility/Need of an SUO

61

5.3

A Standard Kernel Ontology

62

5.4

Concept Maps

66

6

Relati onal DB for Knowledge Representation: Thingism

67

7

Ti me

69

7.1

"The
End of Time" by Julian Barbour

69

8

Irregular Concepts

69

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
3

9

Continuant versus Occurrent and 4D vs. 3D + Time

74

9.1

4
-
D

75

9.2

Continua
nt
-
Occurrent

78

9.3

Change

88

10

Language for the SUO

89

10.1

Ontology: Relation to Natural Languages

100

10.2

The use of English in an
SUO

104

10.3

Intentionality and Cultural Context

104

10.4

Unique Names and Possible Aliases

106

10.5

Logic

107

10.5.1

FOL, or not

107

10.5.2

Irreducible/Essential Triadicity

107

10.5.2.1

Relations and Relations

144

10.5.2.2

Awbrey’s Summary

145

10.5.3

Intension vs. Extens
ion

157

10.5.4

Nth
-
domain

166

10.6

Semiotics

168

10.7

KIF

173

10.7.1

KIF 3.0

174

10.7.2

New KIF

174

10.7.3

SUO
-
KIF

176

10.7.4

Modules, Namespaces, and Importation

176

10.8

EXPRESS and EXIST

177

10.9

EPISTLE

178

1
0.10

OKBC

178

10.11

CYC

179

11

Specific Upper
-
Mi ddle Ontological Concepts

180

11.1

Granularity

180

12

Standard Upper Merged Ontology (SU
MO)
-

Comments

182

13

Courses

217

14

Background Tutorials

218

14.1

References

218

14.2

Awbrey Condensations

219

14.3

Awbrey Instructionals

222

14.4

C. S Pierce

227

14.5

Lambda Calculus

228

14.6

Provability versus Entailment

229

14.7

Foundatio
ns of Math re the SUO

232

14.8

The terms Predicate and Property

234

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
4

14.9

The terms Class and Set

235

14.10

The term Extension

235

14.11

Meaning

235

14.11.1

Lexical and Sentential Meaning

235

14.11.2

Meaning
-
preserving Translations

235

14.11.3

Language Games

236

14.12

Defi
nitions for Algebraic Structures

237

14.13

Concept Analysis

239

15

Ontology Systems and Efforts

240

15.1

CYC

240

15.2

OntoMap

240

15.3

Topic Maps and SERUBA Upper Ontology

241

15.4

WebKB
-
2 and WordNet

243

16

Indi vi dual’s Positions

244

16.1

Matthew West

244

16.2

John Sowa

246

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
5





1

Introduction & Miscellaneous

The purpose and goal of this document is to provide, in a
condensed

form, most of the useful and relevant information made
available via the IEEE SUO email l
ist. There is no editorial summarization:
just filtering/condensation
.


In pursuit of the goal, at most a little violence should have been done to what someone meant or said. The document attempts

to preserve the substance and also the tone of what a co
ntributor said. At the same time, attempts have been made to gently
dissociate an idea from its proponent. So this document is firmly grounded in what happened to have been said via
standard
-
upper
-
ont
ology@ieee.org
. However, it is oriented, as its organization shows, toward the topics and issues relevant to creating a
standard upper
-
level ontology.


Some effort has been made to avoid a contributor’s atypical, non
-
representative statements made in an
“off moment” or made
but later retracted.


Some compromise has been necessary in converting from the message threads to the subject/topic
-
oriented organization of this
document.


Latest Message Included: 13 June 2001



The following “access conrol list
” has guided the inclusion/exclusion of content:

1.

Include material which belongs in standard
-
upper
-
ontology@ieee.org.

2.

Include selected items which belong in ontology@ieee.org and suo
-
kif@ieee.org if they appear to bear on the success of
IEEE SUO.

3.

Exclu
de, in general, items from
suo
-
ce@ieee.org

(Controlled English).

4.

Exclude material on working group procedural isuses, such as voting, rate of progress, and volume
-
of
-
correspondence.

5.

Exclude work
-
a
-
day material on lower

or mid
-
level ontology issues, unless a general principle is at issue. (Simply for
reasons of volume.) For example, exclude the “Axiomatization of Granularity” discussion.

6.

Include material on near
-
term issues while tending to exclude material on “research
” issues not likely to be resolved in the
next few years. (But include assertions that an SUO requires the results of certain research.)

7.

Exclude material that would be unlikely to be understandable after translation into another language.

8.

Exclude material
that is not “on
-
topic” unless it can be extracted from its entanglement with the original topic (Subject
line).

9.

Include other posted material of apparent import.

10.

Exclude everything that didn’t match any of the above.




Bowdlerization complaints are curren
tly received and addressed by
fnc@mitre.org

.





From
http://www.idef.com/idef5.html

:

“Ontological analysis is accomplished by examining the vocabulary that is used to d
iscuss the characteristic objects and
processes that compose the domain, developing rigorous definitions of the basic terms in that vocabulary, and characterizing
the logical connections among those terms. The product of this analysis, an ontology, is a do
main vocabulary complete with a
set of precise definitions, or axioms, that constrain the meanings of the terms sufficiently to enable consistent interpretat
ion of
the data that use that vocabulary.



An ontology includes a catalog of terms u
sed in a domain, the rules governing how those terms can be combined to
make valid statements about situations in that domain, and the sanctioned inferences that can be made when such statements
are used in that domain. In every domain, there are phenome
na that the humans in that domain discriminate as (conceptual or
physical) objects, associations, and situations. Through various language mechanisms, we associate definite descriptors (e.
g.,
names, noun phrases, etc.) to those phenomena. ….”


Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
6


=+=+=+=+
=+=+


Potential "non
-
obviable general obstruction" (NOGO)s:



TIM
) It is necessary that the IEEE SUO be able to represent time.


CHG
) It is necessary that the IEEE SUO be able to represent change.


ACR
) It is necessary that the IEEE SUO be able to represen
t alternative courses of reality.


INV
) It is necessary that the IEEE SUO have no incommensurate views on basic ontological issues.


URV
) It is necessary that the IEEE SUO have no unrelated views on basic ontological issues.


4D
) It is necessary that the

IEEE SUO have a 4D orientation.


NTH
) It is necessary that the IEEE SUO have no n
-
th Domain.


DOI
) It is necessary that the IEEE SUO include a means for articulation of domains of interest for subtending ontologies.


XMP
) It is necessary that the IEEE SU
O be an *example*, or unspecialized prototype, of an upper level
--

which can be
adapted, extended, or partly disabled.


REF
) It is necessary that the IEEE SUO have an integral facility for
reflection

including, minimally, integral capacities for
quota
tion.



=+=+=+=+=+=+

Matthew West:


Other things that would be required are:


-

The scope. What is definitely in, what is definitely out.


-

The architecture. This should talk about and describe the skeleton of the

ontology that is to be fleshed out. e.g
. we might include areas like:


-

Backbone
-

perhaps both a 3D and a 4D backbone with a mapping between them


-

different sorts of whole part


-

elements of set theory, e.g. classification, subclassof, disjoint


-

connections


-

Principles
-

probab
ly just an outline, pointing to a more detailed

methodology document.



=+=+=+=+=+=+

It might be useful to make sure we all have

a shared sense of
the main purposes

to be served, among the several that

have been put forth now and in the past:


search

sens
e disambiguation

inference


within the SUO to how many levels


penetrating into domain ontologies

pedagogy

interlingual core

knowledge discovery


Assuming the SUO to be completed and approved, which of these would it

support and to what extent?


Lee


Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
7

=+
=+=+=+=+=+

ONTOLOGY IS NOT A SEMIOLOGICAL TOPIC
. That is, 'ontology' refers to what there IS, not to how people talk
about it
. It is, right at its very heart, fundamentally, about the world, not about signs; and still less about human *use* of
signs. Pat

Hayes



=+=+=+=+=+=+

it is usually
best to use a sign first

to see what it does, and to
resort to mentioning it only as

a last resort, if it doesn’t do
what one intended.


Pat Hayes


=+=+=+=+=+=+

>(subclass
-
of Schema Abstract)

>(subclass
-
of Schema Contin
uant)


This conflation of 'abstract' and 'continuant' doesn't seem to make sense. An 'abstract' is something which stands outside t
ime:
it has no temporal aspect or dependence at all; it doesn’t even make sense to ask temporal questions about it. (From

the Sowa
explanation: " Forms can be said to exist in the same sense as mathematical objects such as sets and relations, but instanc
es of
forms cannot exist at a particular place and time without some physical encoding or embodiment. Whitehead called
them
"eternal objects" because they are independent of space and time." ) But a continuant is defined to be something that retain
s its
identity during some time
-
interval. (Sowa again: "A continuant is an entity whose identity continues to be recognizabl
e over
some extended interval of time." )


So this notion of 'schema' seems incoherent; and since a lot of rather important stuff is classified under it, none of that
makes
sense either. This seems to me to be the single most central conceptual bug in
the entire scheme, and one that must be repaired

before the ontology can be of any real use.


Similar objections apply to "Script", for the same reason, since

'occurrent' means something that has a temporal location.


>(subclass
-
of Script Abstract)

>(sub
class
-
of Script Occurrent)


I have a similar problem understanding the following categories. In any sense of 'relation' I understand, it doesn’t make se
nse
to say that a relation is either a continuant or an occurent. Relations don't exist through time,

they hold between entities. So
what do these categories even mean?


>(subclass
-
of Structure Relation)

>(subclass
-
of Structure Continuant)

>(subclass
-
of Situation Relation)

>(subclass
-
of Situation Occurrent)

>


=+=+=+=+=+=+

Dear Pat,


> >

> >MW: The real
issue here is with KIF because it does not have

> any mechanisms

> >(I am aware of) for
partitioning an ontology

or of noting

> dependencies

> >between parts.

> > >

>

> Yes, you are correct. KIF is only a language for writing ontological

> content, not

a formalism for manipulating ontologies as objects or

> talking about their structure.
This issue was debated
in the KIF

> working group, and the general consensus was that the language to be

> used for expressing content
-

the logical language
-

shou
ld be kept

> as simple as possible, and the larger tasks of manipulating,

> translating, or partitioning entire ontologies should be left to

> other tools more suited to those specialized tasks. For example, it

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
8

> may be that category
-
theory based ideas

will provide a more useful

> foundation for describing such operations. But in any case, there

> would be no obvious utility in describing those operations in the

> same language used to state the content being operated on. So I don't

> think that thi
s should be seen as an "issue with KIF"; that would be

> like criticizing the wheels of the car for not having a carburettor.


MW: OK but
we need to partition the SUO, and we need to do it now
. We need

some sort of lattice of theories. I am looking for so
me wrapping around KIF

that has this sort of effect:


THEORY my theory;

INCLUDE Basic Theory 1, Basic Theory 2;

....

END THEORY


I agree this isn't pure logic. But we do need something like this.


=+=+=+=+=+=+

>So we need to be able to
define a

>collection

of KIF statements as a theory, and be able to give in a theory

>the list of other theories it is dependent on
.

>

>You do not have any of this structure even informally, so no user would know

>which
bits

of the SUO he needs if say he is just interested in
holes.

>

>If this can already be done with KIF as it is, I would appreciate an

>explanation.


KIF does not of itself provide this structure, but software could be written to extract it automatically. Cycorp have extens
ive
tools to allow the user to 'brow
se' through a large axiomatic Kbase, including such facilities as asking for a concept related to
an existing concept, view and navigate through the (highly complex) class hierarchy, and see all axioms which mention a
concept or group of concepts. It t
akes some getting used to (its considerably more complex than hyperlinks), but one can do
most of what one wants to do in this kind of framework. I honestly don't think we can expect to do much better, and to
develop usable tools like this is not a job
to be undertaken lightly: it is many man
-
years of effort
. No large ontology is going
to fall apart neatly into separate theories which can be taken or left in isolation
.
The SUO isn’t going to have 'bits', it’s
going to be more like a kind of large net
work, with some areas more tangled than others.


Pat Hayes


=+=+=+=+=+=+


1.1

Terminology


=+=+=+=+=+=+

=+=+=+=+=+=+



Jon Awbrey:


The distinction in level of abstraction between token and type is not, in general,

something that we ought to confound with
the distinction in sign
-
relational role between the object and the sign.


Nor is the tone
-
token
-
type dimension relevant to the following absolutely "typical" line of classification in Peirce's theory,

since the tone
-
token
-
type dimension cuts across the met
onymous board of this entire menu:


| Sign

| Index

| Icon

| Symbol

| Term

| Proposition

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
9

| Argument

| Deduction

| Induction

| Abduction


=+=+=+=+=+=+


When you have a word, such as
"class"
, with so many

different meanings to so many different people who are potential users
of the SUO, I believe that it is tainted beyond all usefulness.


Recommendation:



1. Use the word "set" in its purely extensional sense, as in "naive" set theory and its more axiomati
zed versions.



2. Do not use the word "class" as a technical word in KIF or the SUO.



3. Select some term from the set {"type", "sort", "category"), and use it for both the KIF sorts and the things in the


SUO hierarchy.



4. Define a denotation

operator, which for any domain of discourse, maps the things in point #3 to sets.



5. Avoid making a commitment whether the things mentioned in point #3 are or are not identical to their denotations, since
we may want to talk about multiple universes o
r domains of discourse at some time in the future.


John Sowa


=+=+=+=+=+=+

Date: Tue, 15 May 2001 From: "Philip Jackson" <phil.jackson@computer.org>


Adam,


…….I think that in the early stages of ontology development people may need some flexibility….


Besides this flexibility, and openness to different notations, and different

levels of formalization, the other key
requirements
that I would propose for

a "SUO Common Ontology Development Architecture (CODA)" are:


web
-
enabled

open to input from multipl
e authors

an authors input is easily published on the web, directly by the author

readers can view any author's contributions

authors can only edit their own contributions

teams of people can collaborate to develop consensus definitions, or agree

to disagr
ee

multiple ontology efforts / directions can be pursued in parallel

support for namespaces within each ontology effort


To the extent that the
web
-
published database at


http://ontology.teknowledge.com:8080/rsigma/SKB.jsp?req=SC&name=Entity&caseSensitive=on&skb=Merge



meets these requirements, I would say it could be a good alternative to
the SUO Dictionary that I've developed

(with advice
from

Robert Kent).


Phil Jackson


=+=+=+=+=+=+


Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
10

1.1.1

Ontology

I would prefer not to use the word "ontology", except as a buzzword that has become popular in circles that are providing
research funds. Aristotle didn't use it, and the nineteenth century Germans who
made it popular in philosophy were trying to
distinguish theories of "being as such" from the study of beings in the natural sciences.


I regard that move as fundamentally flawed
--

Aristotle properly placed metaphysics after physics, and I don't believe t
hat
ontology can be or should be separated from physics (except as an expedient for keeping any single book from becoming too
big to be held in one hand).


Sowa, 2001/05/01


1.1.2

Context

My preferred definition is given in the CG standard: a concept node that

contains a nested conceptual graph. I am willing to
generalize that term to include any notation for bracketing an excerpt of language (any kind of language) that is under
discussion.


Following is the beginning of my context paper, which discusses the a
mbiguity of the word "context" and my preferred
interpretations: http://www.bestweb.net/~sowa/ontology/contexts.htm


The technical term refers to the purely syntactic function of bracketing some text in a natural or artificial language. The
other,
der
ived uses, are even more important, but I would prefer to use other terms for them.


John Sowa , 2001/05/01

________________________________________________________________________


1. Search for a Theory of Contexts


The notion of context is indispensabl
e for any theory of meaning, but no consensus has been reached about the formal
treatment of context. Some of the conflicting approaches result from an ambiguity in the informal senses of the word.
Dictionaries list two major senses:



* The basic sense i
s a section of linguistic text or discourse that


surrounds some word or phrase of interest.



* The derived sense is a nonlinguistic situation, environment, domain,


setting, background, or milieu that includes some entity, subject,


or topic of in
terest.


The word context may refer to the text, to the information contained in the text, to the thing that the information is about,

or to
the possible uses of the text, the information, or the thing itself. The ambiguity about contexts results from whi
ch of these
aspects happens to be the central focus. These informal senses of the word suggest criteria for distinguishing the formal
functions:



1. Syntax. The syntactic function of context is to group, quote, delimit, or package a section of text.




2. Semantics. The quoted text of a context may describe or refer to some real or hypothetical situation. That nonlinguis
tic
referent, which may also be called the context, constitutes the derived meaning of the term.



3. Pragmatics. The

word "interest", which occurs in both senses of the English definition, suggests some reason or purpose
for distinguishing the section of linguistic text or nonlinguistic situation. That purpose is the pragmatics or the r
eason why
the text

is being quoted.


=+=+=+=+=+=+

Jon Awbrey:


In what sort of conceptual framework could one first begin to carry out comparative and structural, perhaps even dynamic and
developmental studies of different belief systems, knowledge bases, ontologies (with
the sense of information about what is, or
what is supposed to be), mythologies, theories, or whatever you want to call them? This conceptual staging ground would not
yet be a fixed theory itself, but would only be expected to supply the basic equipment f
or investigating various theories that are
set forth. It would need a concept of signs that is general enough to include sentences, like axioms and theorems, but also
far
Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
11

more rudimentary species of expression. It would need a way to talk about sequences
of signs, trajectories through syntactic
spaces, whether these were free associations, streams of reveries, or logical proofs. Within the genera of signs and sign
processes that it gave us the wherewithal to describe, to record, to organize, and to begin
to analyze, there would need to be
room for many different species of semiotic forms of life. But the more regimented and the more routinized forms of logic and

theory would constitute only one peculiar species of formal and normative activity within this
whole realm that I have been
suggesting that we recognize as the world of sign relations.


Sowa: there is such a conceptual framework: Peirce called it
semeiotic
. I highly recommend it.


Among other things,
it includes predicate calculus
, Tarski's model t
heory (which is actually the
Aristotle
-
Ockham
-
Peirce
-
Tarski model theory
), and many
other aspects of sign theory
.


In his paper, entitled "C. S. Peirce and post
-
Tarskian truth," Karl
-
Otto Apel makes it clear that model theory is not a
complete theory of me
aning. But he also makes it clear that it is a necessary, but not sufficient component of any
theory of meaning. (Apel's article is reprinted in E. Freeman, ed., _The Relevance of Charles Peirce_, Monist Library
of Philosophy, LaSalle, IL.)


Peirce likewi
se devoted a great many publications and unpublished manuscripts to various aspects of what is now
called
model theory
. He undoubtedly considered it important, even though he certainly recognized that it was
primarily an example of Secondness
.


But as Pei
rce frequently repeated,
Secondness is a prerequisite to Thirdness
. There is a time for every
-
ness under
heaven.


My own conclusion is that we will not have the setting that we need to compare and contrast different ontologies, in a way th
at
does not for
ce them to compete for living space in our minds, unless we adjoin
that all
-
important third column to our
implicit data bases of objects and signs
, and unless we learn how to use it wisely. This is what I mean by taking
the
interpreter, the participant ob
server
, into account,
and "context" is

yet another, slightly distracting, as it turns out, word for
trying to point to the habitation of this interpretive agent.


1.1.3

Metonymy


“The White House” “Hamlet”, …..


=+=+=+=+=+=+


Pease:



Hamlet the fictional charac
ter



Hamlet an edition of the printed play



A performance of Hamlet



A performance of Hamlet captured on video and encoded as a bit stream



The text of Hamlet as character strings



What Fritz Lehmann has called a "conceptual work"
-

the timeless informational c
ontent of the play



Sowa:



1. The Hamlet example illustrates the pervasiveness of metonymy in ordinary language. It occurs everywhere, in ever
y
discussion from the most mundane to the most abstruse.



2. It is important to recognize it, an
alyze it, and state principles for interpreting it. Ideally, those principles should be
stated in sufficient precision that they can be implemented in automatic recognition procedures. However, doing that

in
full generality is still a maj
or research project.



3. metonymy is so valuable for human communication that I would not recommend that people stop using it. Instead, t
he
burden is on us as logicians, ontologists, and semioticians to formulate principles for interpret
ing it when it occurs (i.e., in
almost every sentence).


Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
12


4. I intend to take advantage of metonymy in everything and anything I write, but I am willing to acknowledge the need

for clarifying my use of metonymy when it is unclear.


John Sow
a


=+=+=+=+=+=+


…… "
metonymy
", which refers to something by
using a term

(often more concrete or "diagrammatic")
to refer to something
abstract
. This usage is common not only in ordinary language, but also in the most formal of all sciences, mathematic
s. We
use terms like "limit", "boundary", or "interval" to refer to numbers, which are the entities denoted by numerals. In fact,
it is
very rare for mathematicians to mention the distinction between numbers and numerals explicitly, unless they are talki
ng about
the actual syntax of decimal, binary, or other representation.


…. use of the word "top" to refer to the most general category of the ontology: it refers explicitly to the place where the m
ark
occurs on the paper or blackboard, by metonymy to the

word instance written in that place, by further metonymy to the word
type, and by further metonymy to the concept expressed by that word.


In programming languages,
a related term is "coercion"
, which refers to the automatic type conversion that takes p
lace when
necessary:



-

Integer to float: In the expression, "2 + 3.75", the


integer value of the numeral "2" is automatically


converted (or "coerced") to float.



-

Character string to numeric: In some languages,


arithmetic can be perform
ed directly on numbers that


are represented by character strings. In "2.6 + '55'"


the string '55' is coerced to the integer 55, which is


then coerced to the floating
-
point value.



metonymy in natural language is extremely common and, I would
say, extremely valuable in general. And I admit that it can
sometimes cause confusion. But I would much rather take advantage of metonymy in what I read, write, and speak than to
force myself and others to insert "conversion" operators for every change
of type.

=+=+=+=+=+=+

Sowa: There has been a lot of good work on distinguishing metonymy such as this, and in a large number of useful cases, it
can be done. Computers are much better than people in detecting ambiguities, even when they cannot resolve
them
--

primarily
because people are so good at recognizing and interpreting metaphors and metonyms that they don't even notice that an
ambiguity exists. [ Awbrey: This is a little bit like saying that folks with weak immune systems are better at detectin
g health
threats, or that canaries are better at detecting toxic gases. [Sowa: Not at all! …. Checking for ambiguities is an ideal
application for computers.]]

=+=+=+=+=+=+

Sowa:

Metonymy and metaphor are very powerful devices for adapting a limited voc
abulary to a potentially unlimited range of
applications. Sometimes, ambiguities may result.


1.1.4

Semiotics, Zoosemiotics, Message

Courtesy Jon Awbrey:


| Semiotics is, quite simply, the exchange of messages.

| A message consists of a sign or a string of si
gns.

| "Zoosemiotics is a term coined in 1963 to delimit that

| segment of the field which focuses on messages given off

| and received by animals, including important components

| of human nonverbal communication, but excluding man's

| language and his se
condary, language
-
derived semiotic

| systems, such as sign language or Morse code."

|

| Thomas A. Sebeok,

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
13

|"Talking with Animals: Zoosemiotics Explained",

| Chapter 6, pages 109
-
116, in:

|'The Play of Musement',

| Indiana University Press, Bloomington, IN,

1981.



1.2

Lexical Items/Terms/…, Polysemy, …

=+=+=+=+=+=+

Date: Sun, 8 Apr 2001 From: Aldo Gangemi
gangemi@saussure.irmkant.rm.cnr.it

by
'intended meaning'

I mean exactly what the 'producer' of
a statement wanted to express in some context for some purpose.
The representation (or sign) relation in the semiotic (or semiologic in saussurean terms) tradition means more or less that
: a
relation between something
-

linguistic item, icon, gesture,
or anything an agent _intends_ to use to express some meaning
-

and a meaning (or conceptualization).

In the recent developments of semiotics, a meaning is
-

to say the least
-

multilayered: for example, for a lexical item
used in a conversation, ther
e is the meaning of the producer, the meaning of the contextual receiver (e.g. someone
participating in a conversation with the producer), the meaning of someone who is not participating in the
conversation but tries to reconstruct it, the meaning of a

text containing the transcription of the conversation for an
average reader sufficiently aware of the topic, the meaning of that text for a reader who is out of topic, or even from
a different historical context or culture, and so on. No surprise that
the main applications of semiotics have been to
literary or art criticism and advertising!


IMO, what computational ontology (as far as lexical items are concerned) is trying to do is formally approximating
the
conceptualization that
an average reader f
rom our current dominant (or shared) culture

assumes

when reading a lexical
item. In other words, only some possible models for a lexical item are committed to in a given ontology.
The more general an
ontology

claims to be, the more it assumes the conc
eptualization of that average reader independently of any context of use.
The more domain
-
oriented

an ontology is, the more it describes the conceptualization of an average reader for a restricted
community or context. In this perspective, intentions a
re not _disturbing_ or _dangerous_ to computational ontology per se,
but they should be _neutralized_ under certain conditions in order to give machines something that they can share with
humans. This kind of investigation has been already formulated by

structural linguists and semioticians at least from the
seventies (ask me the references if you like), and Ferdinand de Saussure made a fundamental distinction between 'parole' (t
he
use of language, carrying intentions), and 'langue' (the language as
system, carrying
shared conceptualizations within a
culture
). Some work was also done, but it was afflicted by some problems: primarily a lack of customers ;
-
), then the sad fact
that those who formulated the methodology of investigation often lacked a
good logico
-
mathematical background, and usually
refused to define something similar to a SUO.


A top level ontology is a good idea also because it is neither to say that the relation between a lexical item and a
conceptualization is one to infinite (t
his is the so
-
called 'arbitrariness' claim, namely a
'radical polysemy' assumption
), but
obviously nor that it is a one to one relation (polysemy is widespread). What the evidence from ecological psychology
(Gibson), cognitive semantics (Lakoff, Talmy,
..), neurolinguistics (Damasio, ..), etc. suggests is that in natural linguistic
systems, such as natural language used by humans in non
-
formally controlled settings, the sign relation is not arbitrarily
applied, but limited by certain cognitive
-
schema
tic invariants. Moreover, formal ontology investigated some areas that can be
used as a formalization of the invariants deriving from the interaction between humans and external world (identity, part,
connection, dependence, unity, etc.).

=+=+=+=+=+=+

D
ate: Sun, 8 Apr 2001 From: John Sowa in response to Aldo Gangemi


> the role of context is in

>disambiguating the lexical meaning of 'have', one of the most

>polysemous verbs in English (WordNet gives 21 entries).


Actually, I [Sowa] believe that Ruhl

makes
a strong case for monosemy

(_On Monosemy_ by Charles Ruhl, SUNY Press,
Albany, 1989). His point is that most, if not all words have a central core meaning and the many meanings listed in a lexicon

can be derived by the techniques of metaphor, metony
my, and specialization of the central core to a subtype that is appropriate
for the language game that is being played in the current context.


The so
-
called polysemy of "have" (or "echein") was noted by Aristotle, but the core meaning of "have" is that th
e subject
stands in a certain kind of dyadic relation to the object. In my KR book, I identified that relation with what Whitehead call
ed a
"prehension". The kind of prehension can be determined from the language game, which is part of the context.


>You
could use a WordNet
-
based ontology library including a definition

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
14

>(more probably, an axiomatic description) of 11 shared average

>conceptualizations of 'little', 4 'of 'lamb', 21 of 'have', some of

><past tense>, some of 'a' (indefinite article). That

is all for

>ontology. If you want to analyze sentential meaning, you have to

>exploit ontologies, but they only provide a network of lexical

>meanings. Then you need a parser, a co
-
occurrence analyzer, an

>abductive inferential engine to check the pos
sible representations of

>the sentential meaning compared to the topic of the text, etc.



<snip>


>
Wittgenstein's Philosophical Investigations

is one of my

>Bildungsroman :
-
)


That should be required reading for anyone who is doing ontology, semantics of

natural language, or knowledge representation
in general. And by the way, in the preface to that book, W. comments on his earlier "mistakes" in the _Tractatus_:


I was helped to realize these mistakes
--

to a degree which I myself am hardly able to est
imate
--

by the criticism
which my ideas encountered from Frank Ramsey, with whom I discussed them in innumerable conversations during
the last two years of his life.


For a study of the historical connections between Peirce, Ramsey, and Wittgenstein,

I recommend


http://www.door.net/arisbe/menu/library/aboutcsp/nubiola/scholar.htm



John Sowa

=+=+=+=+=+=+

1.3

Ontological Entities vs. Lexical Entities and Triadicity

The
other day I reread Chris Swoyer's article on Properties in
the Stanford Encyclopedia of Philosophy



(http://plato.stanford.edu/entries/properties/)


Swoyer uses the term "property" broadly to cover relations as well as monadic relations. I'll stick to
that usage in this message
as well, since I think it will be useful to
distinguish between ontological entities and lexical entities
, and I don't think the
distinction has always been made in this discussion so far. I'll use "property" for the former, an
d "predicate" for the lexical
entity that sometimes corresponds, and avoid the word "relation" altogether.


Reviewing all the many issues concerning properties, including the issue of whether or not they exist, makes me think that
perhaps Matthew has hit t
he nail right on the head with his statement near the end of the message quoted below:


>

>MW: Well this makes me think that you have a different definition

>of what a relation is to me. This would easily explain how we

>have been missing each other.


Pred
icates

of all arities exist in a standard syntax for predicate logic. This doesn't imply that
properties

of any order exist
--

that
depends on your ontology. If I've got this right, one's ontology apparently can depend on
tropes
, for example, without
admi
tting the existence of properties except as bundles of tropes. (I'm not saying I agree with this, I'm only saying that
apparently some philosophers do).


On the other hand, even supposing you believe in properties, whether or not triadic properties are ne
eded may depend on what
kinds of individuals one believes in. To take one of Jon's examples, analysis seems to some people to necessarily be triadic
,
because there is an analisand and at least two analytic components. It seems to me that this depends on
a belief that each
analysis is a triadic or traidic+ property, and thus is a somewhat circular argument.


On the other hand, if you believe that each analysis is an individual (could be process, event, occurent, or ?), then you can

define a monadic predica
te analysis(X) which is true iff X is an instance of analysis. Then, you could believe that there is a
dyadic property, analysand, and another dyadic property analysis_component. This seems to me to cover Jon's example, as
well as allowing for additional

dyadic properties such as analysis_performer, analysis_epoch, and so on. The essential three
-
ness arises, of course, from the understand of analysis as a 'taking apart', and would be represented by
axioms that say
Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
15

something like "X is an analysis only if

(1) there exists analysand of X, and (2) there exist at least two analysis_components of
X.

'giving' and so on could be treated similarly.


It seems to me that the <emph> engineering issue </emph> is whether dyadic predicates can communicate an intended
interpretation that is satisfactory to those who believe that the real underlying property is triadic.


Best regards,


John Velman

=+=+=+=+=+=+

=+=+=+=+=+=+

=+=+=+=+=+=+



1.4

2001, March 6
--

The Story So Far


Date: Tue, 6 Mar 2001 10:09:27 +0100

From: "Wes
t, Matthew MR SSI
-
GREA
-
UK" <Matthew.R.West@is.shell.com>


Dear Colleagues, ....


There are a number (say about 4
-
5) of world views that are documented, either in natural language, or in a more formal form,
with a number of possible variations on these.


Ia
n and his supporters argue that because all philosophers do not agree on just one of these, we should ignore all of them,
giving preference to a process that Ian might describe as pragmatic selection from various sources and merge.


I, Pat Hayes, John Sowa
, Chris Partridge, Nicola Guarino (please correct me if I am mistaken) and perhaps others would argue
that we would be better served by understanding the existing World Viewpoints and relating them to each other.


Discussion


As far as I can see there are

four possible outcomes to Ian's approach
.


1. Ian solves the problem that previous philosophers have failed to solve and creates a single ontology that everyone agrees
is
how the world is.


2. Ian recreates one of the possibilities that are already known
about.


3. Ian creates a new ontology with a different world viewpoint than those already existing, adding one more to the list that
philosophers don't agree about.


4. Ian fails to create a consistent ontology.


Let us consider these in turn:


1. Included

for completeness only. I don't think Ian expects this outcome.


2. One of the two most likely outcomes, in which case it would be more efficient to do some homework and make an informed
choice (or adopt the counter position above).


3. A remote possibilit
y, and probably the least useful.


4. The most likely short term result, based on what I read and my own experience, getting to a consistent universal ontology
from scratch takes some 5
-
10+ years work
--

for those few who can claim some measure of

success.


The alternative approach:



-

Recognises that there are a (small) number
of major world viewpoints

that exist and are valid (within some range).

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
16


-

Tries to understand explicitly what are the elements that underpin those viewpoints and the applicable ran
ge.


-

Identifies key choices that are mutually exclusive.


-

Documents those viewpoints based on that understanding.


-

Enables interoperation between those viewpoints by


mapping between them.


We could of course pursue both of these options.


Comments
?


Regards


Matthew


=+=+=+=+=+=+

Awbrey:

the use of the word "module" in mathematics is subject to the following analogy:

Module : Vectorspace :: Ring : Field.


=+=+=+=+=+=+


1.5

2001, May 22


Now what?

James Schoening wrote:

>

> All,

>

> Thank you
for your votes.
Now what?

>

>
As I see it, we need to start working on one or more documents

> to have any chance for success. As such, I have contacted those

> with a candidate starter document, to see if they have given thought

> to submitting it as a

group work effort. If I overlooked you, please

> let me know.

>

> Until then, let's try to keep some degree of fucus in discussions.

>

> Jim Schoening


¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤


Jim, & All,


I need to try and articulate for t
he group, in some way or another, a problem that I have had with the way of working that you
suggest, and perhaps you or somebody can clarify for me
why it is so taken for granted that there can be no other way
. I am
speaking about the process of focusing

on a candidate document, and somehow incrementally bootstrapping it up to a real live
ontology.


In my way of looking at it, an ontology is like a single data structure, perhaps an implicit or a virtual one, but still just

a one
-
off
schematic. Every bit o
f training that I have ever had, whether in math or computing, tells me to go for the generic process
instead of the isolated structure. I can almost understand the utility of tailoring domain ontologies to the specific domain,

but
even there,
every signif
icant domain that I have ever been acquainted with long ago passed the stage of complexity and
difficulty where any sensible community of researchers would even want to think about building an ontology for it by hand
, the
way that we used to attempt to do
way back in the naive glory days of "expert systems". All of this leads me to the inexorable
conclusion that it is just plain bad methodology to work on or toward a single, hand
-
made data structure without putting an
effort into
the programs that would ha
ve to maintain it, both to create it from rawer data
--

which is where our axioms,
concepts, definitions, and taxonomies ultimately come from, or at least have to be tested on
--

and also the software to check it
logically and draw inferences from it.
That
's about as plain as I can make it. It has always seemed so obvious to me, as
something that I just took it for granted that everyone already understood, that I still feel just a little embarrassed to ha
ve to say
it.


Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
17

When I reflect on this impasse, it oc
curs to me that maybe different people just have different ways of working. I got trained in
proving theorems and solving problems. I often try to treat this as the kind of problem
-
solving seminars that we used to have
in grad school. Maybe that's the w
rong model, but it's one of the best
models of collaborative work

that I have in memory. But
I get a little unnerved, though, when people pose what they describe as a "problem"
--

I almost reflexively go to work on it
--

but then it quickly becomes clear t
hat they are literally "filtering out" any brand of solution but the one that they already had in
the back of their heads going into it, and this tends to happen a lot here.


Anyway, just in case there happen to be other members of the group that share thi
s attitude toward the work, let me propose
that we establish an alternate or parallel track, one that focuses on "focal" problems, that is, moderately well
-
defined or sharply
outlined obstacles to progress in building ontologies or the software for generat
ing and maintaining them.


Sincerely Yours,


Jon Awbrey


2

Ontology: The State of the Art

Date: Wed, 16 May 2001 From: David Whitten <whitten@lynx.eaze.net>


I would like to know something about the current state of the art

in ontology practice.


>

>

> >John F. Sowa said:

> >

> >I would like to address the concluding paragraph of one of your

> >recent messages:

> >

> >pat hayes wrote:

> >

> > > If you want to be able to talk about what might be true, then use a

> > > modal logic explicitly, and keep

that part of the ontology sharply

> > > distinct from the parts that do not involve modal notions.

> > > Incorporating modality into the structural ontology would make the

> > > SUO worthless for all industrial and engineering applications.

> >

> >John F.

Sowa replied:

> >I agree that for most practical applications, a modal theorem prover

> >is highly undesirable. However, modality does enter into commercial

> >database systems at the metalevel (although that point is seldom

> >recognized by the people w
ho routinely use those features).

>

> pat hayes responded:

> OK, and thanks for the exposition. But I would just emphasise that

> this is the METAlevel we are talking about here. I agree with your

> point that updating and modalities have an interesting

correspondence

> here. At the metalevel, of course, we theoreticians are free to use

> whatever mathematical tools we feel like doing.

>


Okay,

do any robust modal theorem provers actually exist today in May, 2001, such as are referenced by John?


Hay
es: There may be some, but I know of none. A modal theorem
-
prover would be functionally indistinguishable
from a first
-
order TP applied to a sitation
-
calculus style transliteration from the modal langauge back into FOL.
What John is talking about is som
ething that reasaons at the metalevel of the FOL reasoner, however, and this seems
like a much more ambitious notion, rather along the lines of McCarthy's old dream of an 'advice taker' program.


Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
18



Do any meta
-
level description systems exist that implem
ent the mathematical tools Pat is speaking of?


Hayes: I am not the person to answer this. I think John's point is that there are the tools, and then there is the
modal/meta account of them , and that the people with the tools arent modal/meta savvy eno
ugh yet to see the
connections.


Is this discussion primarily about ways of viewing possible ontological design?


Or are these tools being actively used by practitioners in knowledge base systems today in May 2001 ?


Can the SUO working group use any of
the tools (that could) exist that allow for meta level operations and modal viewpoints ?


What time frame of availability are we discussing?


What type of training/education is necessary to be effective in using these physical programs, or mental mathemati
cal tools?


David



3

Ontology Development Methodology

Date: Fri, 13 Apr 2001 From: "John F. Sowa" <sowa@bestweb.net>



Pat and Robert,


Your discussion reminds me of a point that I believe we should highlight in the tutorials or commentary about the me
thodology
for developing ontologies (or any formal theory of any kind).


>Again, I would say not, in general. The basic point is that it is

>impossible to say that a particular ontology *completely defines* the

>meaning of a symbol. If we could say in so
me sharp way that A

>completely defined the meaning of r, while B and C only used that

>meaning, then I would agree that it would be otiose to insist on

>distinguishing b1#ai#r from b2#a2#r; but I don't really believe this

>ever happens; and certainly
there is no model
-
theoretic basis for

>every saying such a thing, since every assertion made using 'r' adds

>some component of meaning to it by restricting its possible

>interpretations in some way...


I agree. But we should include a comment about
dif
ferent kinds of definitions
. I suggest some discussion and examples,
such as



1.
Closed
-
form definitions

that define one symbol or


expression as an exact synonym for another; e.g.,



f = (lambda x,y)(2xy
-
17)



2.
Axiomatic specifications in wh
ich some symbol is


introduced as a new primitive

that participates


in certain relationships with other incompletely


defined primitives; e.g.,



Monadic predicates: dog, cat, animal.



Axioms:



(Ax)(dog(x)
-
> animal(x))

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
19




(Ax)(cat(x)
-
> animal(x))



~(Ex)(dog(x) & cat(x))



These three axioms tell us that cats and dogs are animals


and nothing is both a cat and a dog. But the meaning of


these three predicates will continue to be specified in



more detail as new axioms are imported from other modules.


>Another possibility is to allow an ontology to import another one

>without any prefixing. If this option were generally used, of course,

>all these identities would be automatic. There would h
owever be a

>great danger of inadvertent inconsistencies appearing.


Sometimes we want to do that. As in the example about cats and dogs, we may want to import information about them from
many different sources in order to get a fuller picture of what it

means to be a cat or a dog.


John



=+=+=+=+=+=+

=+=+=+=+=+=+

=+=+=+=+=+=+




From: pat hayes <phayes@ai.uwf.edu> To: Adam Pease
apease@ks.teknowledge.com


………….

…………

Has it not occurred to you that
this may simply be a sign that Lenat and Hendler might be right, and that the upper ontology is
largely irrelevant to the lower ontological levels? At the very least, does it not suggest that to get the easier (lower) p
arts done
first ( particularly wh
en so many of them are already done in some detail), and then to abstract the upper levels from them,
might be a useful approach, rather than insisting on this top
-
down methodology? It seems kind of obvious to me that it almost
impossible to get the up
per levels right in isolation, and it is harder to do that than to abstract them, so why bother?


……………………………………………

>A great deal of misunderstanding happens I feel because we're

>talking in ambiguous English. Once people get to the point of being

>wil
ling to state theories as formal axioms, I think resolution on

>many issues will be quicker.


I think you are confusing genuine disagreement about substantive issues with the kind of misunderstandings which arise from
ambiguity. Moving to a formal langu
age does not produce instant clarity, since the intended meanings of the formal assertions
have to be elucidated in order that the formalism can be judged. One could, admittedly, take a more rigorously logical line

towards this issue, and insist that a
xioms say what they say, regardless of human intentions (a position I myself am in
sympathy with, as I expect you know.) But
*that* position requires its own kind of rigor: one would then want to see
demonstrations of the consistency of the ontology, by

actual descriptions of models, and some effort made to locate
nonstandard models

(to see the kind of thing I have in mind, take a look at the comments on models in my time catalog, which
were inspired by Van Benthem's work.)

…………………………………..

>I've neve
r said that people can't hold different points of view,

>just that compromise and agreement is possible.


How would you compromise between a view of time which says that time is discrete, and view of time which says that time is
dense? That is
like comp
romising between P and (not P)
. Yet both views have their uses and their adherents. Similarly for
space being continuous but surfaces being real; for time being linear or branching; for things seen as continuants or as
processes; for spatial properties
seen as applying to locations or to things; for temporal properties being inherited by
subintervals or applied to reference intervals; for fluid objects being coextensive with pieces of fluid or not; and so on
and on
Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
20

and on. Almost everywhere I look at

any ontological issue, I can see alternative, incompatible but both potentially useful ways
of describing it.

From: jeff
jeff@polexis.com


If this is right, as I think it is, then it seems to call for a bit of

re
-
engineering. The ontology should be able to contain
the things the things that reasonable people (or their agents) can put together into these sorts of "alternative,
incompatible both both potentially useful" points of view, and the reasoning framework

should be able to
accommodate such incompatibility. Such frameworks exist, I have in mind SMODELS or DLV.


SMODELS: http://www.tcs.hut.fi/Software/smodels/

DLV: http://www.dbai.tuwien.ac.at/proj/dlv/

See also:
http://www.cs.engr.uky.edu/~mirek/computing
-
with
-
dl.ht ml



Jeff Lansing

These are all expressively limited logics which are

designed for efficient reasoning and default handling. The

compatibility issues we are facing ru
n deeper than anything these

mechanisms can handle, I believe.

Pat Hayes

………………………….


Pat


PS. I have been browsing
the merged ontology

and so far *every single thing* I have seen in it
is wrong
. In particular (in no
particular order):


1. the axiom for

piece
-
of is obviously wrong, since it entails that every piece has all the properties of the whole: (=> (piece
-
of
?X ?Y) (forall (?Z) (=> (instance
-
of ?Y ?Z) (instance
-
of ?X ?Z))))


Ian Niles: I don't see why this strikes you as obviously wrong. I def
ined 'piece
-
of' to be a specialization of 'part
-
of',
where both of the arguments are stuff
-
like entities. You may believe that such a specialization is not useful or you
may be violating your own dictum that the meaning of a term in a formal language is d
etermined solely by the axioms
that contain it. It's hard to see what you're up to here.


Pat Hayes: No, my point is that this axiom says that every property (forall (?Z)...) of a piece of stuff is
inherited by all its pieces. That is obviously false. (F
or example, the
pieces typically weigh less than the
whole
; obvious things like that.) If you mean this to apply only to certain 'relevant' properties, then you need
to restrict that quantifier to include a condition like (relevant ?Z), and then give us so
me axioms for relevant.



Ian Niles: No,
actually what
you

say is false
. According to the merged ontology,
the second
argument

of 'instance
-
of'
is

required to be
a class
, i.e. a natural kind, and the examples of
"irrelevant" properties that you cite simpl
y don't and wouldn't show up as classes in the ontology.


Pat Hayes: Then (1) you need to reconcile this usage with Chris Menzel's

structural ontology, where 'class' is synonymous with 'property' and

(instance ?y ?z) is equivalent to (?z ?y);

(2) you nee
d to axiomatize 'natural kind' (and good luck, since one

definition of 'natural kind' is that it refers to any concept that

cannot be fully axiomatised); and

(3) my point still applies in any case, since it is just as true of

natural kinds that they are

not all inherited under parthood.


Look, I know what your axiom is trying to say. What you want to say is that every part is
similar to every other part in *certain ways* , where 'certain ways' refers to a certain
category of properties that are the same
of all parts of a 'uniform' thing. Properties such as
being made of a certain stuff, being at a certain temperature, and so on. The trouble is that
there are plenty of other properties that do not characterise these 'certain ways', and do not
satisfy this
axiom. Until you nail down the properties you want to talk about, this axiom is
either false (as it stands) or useless (if restricted to 'relevant', since that can only be defined
Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
21

so as to make the axiom circular. on the face of it.) Your notion of releva
nce is trying to
prop itself up, but it can't stand on its own.




Chris Partridge: I do not know whether I am interpreting you (or Pat) correctly, but you seem to be
missing a well
-
known aspect of stuff
-
like entities here (and thus Pat's original comment
s about
being wrong are correct
-

though maybe not 'obviously wrong' to everyone).


Typically a stuff's properties (as stuff) are only inherited down to a certain degree of granularity
. I
seem to recall raising this point some while ago, when mass and coun
t were discussed. As I recall
Nelson Goodman calls this property
dissective

-

in Structure of Appearance. You can also find it
described in many places (as well as on this list)
-

for example, I seem to recall it mentioned in Aldo
Gangemi's Onions ontology
.


On a more 'sophisticated' note
you

also
seem to be

implicitly
assuming

a kind of
essentialist/continuant ontology
-

where there is
a sharp ontological distinction between mass and
form (?)

-

where a piece of wood cannot be a stick (or a piece of clay ca
nnot be a statue
-

otherwise,
by your axioms, all its parts would also be statues).
This

is the kind of implicit assumption that
bedevils any attempt to construct an ontology

and as I have argued elsewhere
it would make sense
to make these kind of implicit

assumptions explicit
-

both to help in the construction and the
assessment of the ontology
.


2. piece
-
of only applies to ContinuousObjects, which is wrong.


Ian Niles: See my comment above

Pat Hayes:
OK
,
but

then you should
document somewhere that piece
-
of doesn’t mean "piece of".


3. OrganicSubstance is a subclass of ContinuousObject, which is either wrong or assumes a strongly mereological view of
objecthood.


Ian Niles: I don't follow you here. All I meant by OrganicSubstance is a ContinuousObject
which is also organic. Of
course, you had no way of knowing this, since I neglected to include any documentation for this concept.


Pat Hayes: I was reacting to
the idea of putting (any kind of) Substance as a subclass of (any kind of) Object
.
The only wa
y I can make sense of that is as a declaration that
you are thinking of objects as mereological
sums of substance
: 'Pieces of thing
-
stuff' rather than 'thing'. You need to get clear on what your attitude is
towards the
thing/substance distinction
. (You m
ight want to reject it, which is fair enough as long as you face
up to the ontological consequences of doing that.)


4. The
explanation

for continuousObject does not make sense. It starts "A ContinuousObject is an object in which every part is
similar to

every other in every relevant respect." which is a contradiction in itself, since some parts presumably contain other
parts as parts.


Ian Niles: This would be a contradiction only if one assumed that the relevant respects apply to all of the parts of
the
original parts, and I am not making this assumption. Generally speaking, when one distinguishes stuffs from objects,
one claims that the stuffs are stuff
-
like only at or above a certain level of granularity, and, hence, parts of things at this
level o
f granularity will not be stuff
-
like.


Pat Hayes: But that is not what the text I cited says. It says "..in which EVERY part is similar...". My point
wasn’t about the granularity issue, but about the fact that a part is not similar to the whole of which

it is a
part in every respect.


Ian:: I just don't follow you. The prepositional phrase "in every relevant respect" was intended to
modify "similar" in the original docu
-
string. In any case, enough said, and I'll try to write more
clearly in the futur
e.


By the way, that idea of
"down to a certain level of granularity"

is a fine aspiration for an ontologist, and
everyone sees that we need it,
but nobody has actually managed to do it
. I don't know of anyone who has
axiomatized the idea satisfactorily

(yet). The nearest is Jerry Hobbs, and he had to hack round the basic
Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
22

problem by assuming that approximation is transitive (which of course it isn't, as every engineer, builder
and carpenter knows in detail.) I've been trying to do this, on and off, f
or about 30 years and I haven’t
managed it yet, but I do have a very finely honed sense of how not to do it and why all extant ideas don't
work.


The central problem is that
the heap paradox

has a very nasty interaction with logical inference: it
violat
es
the transitivity of correctness
. About every decade or so someone publishes a paper about granularity or
approximation which fails to solve the basic problem.


5. Can something be a selfConnectedObject and also a ContinuousObject? If so, what is a
superficial
-
part
-
of a
ContinuousObject, and if a superficial
-
part
-
of is also a part, is it similar in every relevant respect to a non
-
superficial part? If
not, why is the object Continuous; if so, what makes the part superficial?


Ian Niles: Yes, the m
erged ontology allows that a ContinuousObject can also be a SelfConnectedObject. A
superficial part of a ContinuousObject would be a part of the ContinuousObject at the surface which has the precise

granularity for the stuff.


Pat: You are relying on th
is notion of granularity

to do a lot of your work here.
I would urge caution
, since
you will have a very hard time cashing out this intellectual loan.


Ian: True, I am relying heavily on the notion of granularity and I would have a

difficult time fully ax
iomatizing this notion, I'm sure. However,
I don't

see that we're required to axiomatize this notion
. Of course, it would be

nice if we could specify this notion with necessary and sufficient

conditions,


Pat: I said nothing about necessary and sufficie
nt conditions. (You might

read my old papers on naive physics to see my opinions on the

irrelevance of giving such definitions.) But it is irresponsible to

say that you are not going to give *any* account of a concept which

is notoriously problematic a
nd which you also propose to give a

central role in an ontology. Reasoners are going to have to be able

to draw *some* conclusions.


but
there are doubtless going to be some primitive concepts in

the ontology, and granularity looks to me to be a good can
didate

for this

category.


Pat: All concepts are primitive. We still need axioms about them.


By the way, what do you mean that the surface has a granularity for the stuff? Surfaces are not made of stuff,
surely (if they were, they would have properties
like mass and density.) Or do you mean, the stuff of which
the object is composed? If so, I fail to see quite what that has to do with the notion of surface. You need to
be very careful about topology here. What do you mean by a part being AT a surface
? The part can't be
literally inside the surface or literally share the surface's location, since you can't get a 3
-
d thing (even a
very thin one) inside a 2
-
d thing. (I'm assuming here that the surface itself isn't the part; if it is, then there are
ot
her, different problems.) Have you tried stating this condition carefully?


A superficial part could be similar in every relevant respect to a non
-
superficial part, because I think we're all agreed
that relevant respects do not encompass locative properti
es.


Pat Hayes: We are? I was going to ask you what 'relevant respect' meant, in fact, and how you were going to
axiomatize it. Maybe you could expand on this point a little.


Niles: My response is just what I said above. Since 'instance
-
of' requires
a class as its second
argument, we don't have to worry about the sort of silly properties that seem to concern you.


Hayes: Maybe you should read up on "class". A class is simply a property. Classes have
no magical properties of 'naturalness' and are not

guaranteed to not be 'silly', unless you
Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
23

somehow provide axioms to rule out the silliness. (For example, would having a certain
mass be a silly property?)


Finally, part of what makes a part superficial is that it is located at the surface of the object.


That doesn’t quite make sense, since a part is 3
-
d and a surface is 2
-
d. What you might mean (?) is that the
part is in contact with the surface, or that part of its surface is part of the surface of the whole thing of which
it is a part, i.e. it share
s some of the whole thing's surface. But even that will not do, presumably, since that
would also apply to the whole object. I really don't see any way to characterize 'superficial part' in terms of
location alone.


Ian: I agree. I think I spoke too lo
osely. Casati and Varzi, from whose book the notion is taken,
characterize a "superficial part" as a part which does not have an "interior part".


Hayes: OK. I see how they are doing it. You might want to check carefully however on
whether 'superficial p
art' means something like a surface or something like a skin, because
the things you say about it elsewhere will be different in the two cases.

.....


6. An
action

is a subclass of a HumanCaused Process, which makes actions by animals, computers and other
machines
logically impossible.


Ian Niles: It all depends on how you define the notion of 'Action'. One can draw a distinction between events that are
intentional and the result of deliberation and events that are not. This sort of distinction, you are
right, leaves (most)
animals, robots, etc. out of the picture.


Pat Hayes: No, it does not. Certainly many animals, and arguably also some machines, can act as the result
of deliberation and with intention.


This is a highly contentious area of academi
c debate, as I expect you know. It is inappropriate for a standard
ontology to take a stance (particularly such a narrow
-
minded and rather archaic stance) on such an issue,
and have this stance welded into its categorization hierarchy.


This alone woul
d be sufficient for me to refuse to ever use the ontology and to recommend that it never be
used in any scientific context. I mean, ask yourself. Here I am in AI and cognitive science, and I'm expected
to agree to using an ontology which is built on the

assumption that the basic scientific hypotheses of my
chosen field are false *by definition* ?? You might as well ask an MD to use an ontology which is based on
the 4 humors. (I actually feel rather strongly about this because I think that the idea tha
t only human beings
have intentionality is close to immoral; but my point is not that you should agree with me on this, but that
the SUO shouldn’t have such narrow
-
minded prejudices built into it.)


Ian Niles: OK, how about we replace the term "HumanCau
sedProcess" with
"IntentionallyCausedProcess" and revise the corresponding axiom accordingly. How does that
sound?


Hayes: Much better, thanks.


However, it is a crucial concept
-

in fact, our legal system is founded on it.


Hayes:
Then put it into a le
gal ontology
, not a general
-
purpose one. Ethology, psychology, and much of
biology and philosophy explicitly denies it, and
a denial of it is a cornerstone of cognitive science
.


If one wants to create a more general concept that includes both 'Action', a
s currently defined, and the things that
animals and robots do, this is possible in the framework of the merged ontology.


Hayes: That is not an adequate response. In this sense anything is possible. One could simply ignore the
categories in the ontology

and invent ones own.
The issue is the extent to which the SUO reflects
reasonably widely accepted categorizations.

If "animal" and "human" and "action" and so on have arbitrary
definitions that reflect a particular narrow view of the world which is i
nconsistent with large areas of current
science, then there is something badly wrong with it as a proposed SUO.

Las Saved at
11/1/2013 10:37:00 AM


Printed at
11/01/13

10:37 AM

Page
24


Niles: Well, we haven't yet taken up the definitions of 'Animal' and 'Human'. Why don't you have a
look at them and see if they're to your l
iking. As for 'Action', I suggest that we make it a subclass of
the newly minted class 'IntentionallyCausedProcess'.


=+=+=+=+=+=+

=+=+=+=+=+=+

=+=+=+=+=+=+

Date: Tue, 15 May 2001 From: "John F. Sowa" <sowa@bestweb.net>


…Having ……. only **one** midd
le ontology ……is perhaps the most dangerous possible situation that the SUO could get
into. I believe that
we should concentrate more on a methodology for creating arbitrarily many middle levels (and even
top levels) that are tailored for different purpos
es.


That is possible if we consider the distinctions to be more fundamental than the categories. Following is the general outlin
e:



1. For each distinction (which may be binary, ternary, or a spectrum


such as colors), there are axioms that character
ize each of the


possible options. (A spectrum would be characterized by axioms


that contain a continuous parameter, such as wave length.)



2. The categories of the ontology are defined by conjunctions of


distinctions, and their corresponding
axioms are the conjunctions


of the axioms associated with each of their distinctions.



3. A Leibniz
-
style lattice, which permits arbitrary conjunctions of


distinctions, generates extremely bushy lattices with many nodes


that are never populate
d with instances of possible existents.


But other lattice
-
generating techniques, such as
FCA
, accommodate


constraints among the distinctions that prune the unpopulated


nodes.



4. At any time, new distinctions may be added or deleted, and a ne
w


lattice can be generated by pushing a button.


For a brief summary of lattices with a discussion of the Leibniz
-
style and FCA
-
style, see Section 7 of my tutorial on math &
logic: http://www.bestweb.net/~sowa/misc/mathw.htm#Lattice


John Sowa


=+
=+=+=+=+=+

Date: Thu, 24 May 2001 From: Patrick Cassidy <cassidy@micra.com>


……

> But why do we
need

such a concept than "generic defining characteristics of a physical object"?

>


I think that, in building an SUO,
it is a good idea to include all th
e meaningful higher
-
level concepts that one can
,
because

it helps in maintaining a complex ontology to have each meaningful distinction defined at one high
-
level point rather than at
multiple lower
-
level points. Also, I think it helps to fully grasp wha
t the concepts mean (for me, anyway); and in the event
that someone in the future wanted to create and use some high
-
level concept that is not already in the SUO, it is likely to be
more troublesome at that time than to have it defined at the earliest

point in the process. We needn't rack our brains to think
of all possibiities, or to automatically create all possible concepts, but where one seems natural, I think we would need to
have
a specific reason not to include it. A high
-
level "physical obj
ect" otherwise unspecified is one that is very common in
ontologies, and we should try to define it in a way that is most useful and most easily understood by all users.


Bernard Vatant: Agreed. The difficulty is to be as exhaustive as possible while ke
eping some consistency. The most
"natural" concepts are the ones which turn around to be the most fuzzy and tricky to define when you begin to think
about them ... Following Saint
-
Augustine: "When you don't ask, I know what it is. When you ask, I don't kno
w any
more" :)



Pat Cassidy

Las Saved at
11/1/2013 10:37:00 AM