SUO: RE: Foundations for ontology
John,
. In my previous response I inadvertently omitted the word
"more".
. I have had a chance to go through the presentations at home
last night.
. I see a major shortcoming in the lack of preparedness to
accommodate more sophisticated meaning analyses and interpretations. In
particular, most of the examples of multiple meanings you quoted amounted to
metaphorical ones.
. I realise that having an information processing paradigm
that separates out the metaphorical applications implies processing, rather
than just database access. Nevertheless, it greatly simplifies the basic
vocabulary. With computer power ever expanding, I suggest that processing
can be accommodated.
. The advantage is that most word have only a single basic
meaning, and the application of metaphor is then assessed separately.
Furthermore, I suggest there aren't overly many common metaphorical
transmutations.
. This is a bit like my criticism of your examples of
conceptual graphs many months ago, whereby I felt you weren't breaking the
options down by fundamental and orthogonal categories. If one does apply
such information management approaches, the task is significantly
simplified, though it does require assessment by logical thinkers, rather
than just anybody. (I have included my email on this below yours). I feel
that in part, because the CG I commented on then is the same as in your two
latest papers.
. Furthermore, I suggest orthogonal analysis, including
separation of basic and metaphorical meanings, lends itself to automation.
. I feel significant leverage is being lost through lack of a
willingness to undertake this sort of systematic analysis.
. For you to ponder and possibly respond.
Cheers Graham Horn
National Data Standards Unit
Australian Institute of Health and Welfare
================================================
Phone: 02.6244.1094
Fax: 02.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: Horn, Graham
Sent: Monday, 13 August 2001 14:18
To: 'John F. Sowa'; standard-upper-ontology@ieee.org; cg@cs.uah.edu
Cc: phayes@ai.uwf.edu
Subject: RE: Foundations for ontology
John,
. I wish I could spend time digesting the work you have done.
It is evidently significant and perspicacious.
. Might I suggest you put a motion to the SUO group (or a
modification to an existing one) for us to consider.
. I suspect you are bringing issues to the forum that
transcend the considerations most of us have made to date.
. I believe we would be well served by considering our way
ahead in terms of the issues you have raised, as well as considering your
recommendations.
. What do others think?
Cheers Graham Horn
National Data Standards Unit
Australian Institute of Health and Welfare
================================================
Phone: 02.6244.1094
Fax: 02.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: John F. Sowa [mailto:sowa@bestweb.net]
Sent: Monday, 13 August 2001 13:42
To: standard-upper-ontology@ieee.org; cg@cs.uah.edu
Cc: phayes@ai.uwf.edu
Subject: SUO: Foundations for ontology
As I have said in many notes to SUO list, I have some concerns about any
ontology that is developed by hand. In two recent talks, I presented my
views of how ontologies should be developed. The first talk, which I
presented at ICCS on August 3, surveys the philosophical foundations.
Following are the slides:
http://www.jfsowa.com/pubs/signtalk.htm
The second talk, which I presented at an IJCAI workshop on knowledge
discovery on August 6, suggests automated or semiautomated methods of aiding
in ontology development. Following are the slides for that talk:
http://www.jfsowa.com/pubs/autotalk.htm
I had originally intended to make this one talk, but as it kept getting
longer, I split it in two. There will also be a paper, which I'll announce
later (when it is finished).
At IJCAI, I also attended a talk by Stuart Shapiro. The paper for that talk
included some comments about Cyc, which are very relevant to ontology
development by hand. The problems with keeping Cyc consistent indicate why
I believe that more autotmated and semiautomated tools are essential for
ontology development. I scanned the paragraphs about Cyc from the paper and
included them at the end of this note.
As I said at the SUO workshop at IJCAI, I believe that something along the
lines that Robert Kent has been proposing for IFF is a necessary component
of any ontology project. I would prefer to see SUMO split up into multiple
smaller theories that could be combined by belief revision methods.
John Sowa
Some Observations about Cyc
[The following comments on Cyc have been extracted from a paper that was
presented by Stuart Shapiro at an IJCAI Workshop (citation below). The
evaluation of Cyc is based on Cycorp documentation and on experience by the
first author (Frances Johnson) during a Cyc training course.]
Doug Lenat and Cycorp have developed Cyc [Cycorp, 200la] -- a large
knowledge base and inferencing system that is built upon a core of over a
million hand-entered assertions or rules about the world and how it works.
This system attempts to perform commonsense reasoning with the help of this
large corpus of beliefs (mostly default with some that are monotonic). It
divides its knowledge base into smaller contexts called microtheories which
contain specialized information regarding specific areas (such as troop
movement, physics, movies, etc.). Belief revision is performed within
microtheories or within a small group of microtheories that are working
together, and the system is only concerned with maintaining consistency
within that small group (as opposed to across the entire belief space). For
example: in an everyday context, a table is solid, but within a physics
context, it is mostly space (between atoms).
A belief can have only one truth value, so no microtheory can contain both p
and ~p. For example, ~p could be expressed as the proposition p with a
truth value of false. The technique for maintaining consistency is to check
for contradictory arguments whenever a proposition is derived or asserted
into a microtheory. When contradictions are found, their arguments are
analyzed, and a decision is made regarding the truth value of the
propositions involved. Rankings of beliefs, however, is not a part of the
system -- it uses specificity to determine the truth value of a default
belief. For example: Opus the penguin does not fly, even though he is a
bird, because penguins don't fly. If there can be no decision based on
specificity, the truth value of the default belief is unknown. A default
belief loses out to a monotonic one. And, lastly, according to Cyc trainers
and other contacts, contradictions that are purely monotonic bring the
system to a halt until they are fixed. During Cyc training, Johnson
attempted to prove this last statement and failed -- revision was performed.
The instructors were surprised, but thought the training interface might be
the cause. We plan to explore this further, but it is an example of a
system behaving differently than it is described.
As mentioned [above], Cyc did not perform as described, and there must be
some question as to other possible differences from design theory. Most
specifically, Cyc literature [Cycorp, 2001b] claims to keep the
microtheories consistent, for lack of a better word. When asked, contacts
agreed that, in spite of a cursory check, it was possible that unknown
contradictions might exist that had not, yet, been derived. In this sense,
Cyc can only guarantee that its microtheories are not known to be
inconsistent (or KS-consistent). Ideal terminology, such as consistent and
derivable, is often not appropriate for use with a large or complex
implemented system.
References
Cycorp [2001a] _Cycorp, Creators of the Cyc Knowledge Base_,
http://cyc.com
Cycorp [2001b] _Features of CycL_, http://cyc.com/cycl.html
The original article from which these paragraphs were extracted:
Frances L. Johnson and Stuart C. Shapiro, "Redefining belief change
terminology for implemented systems," _Inconsistency in Data and Knowledge_,
Working Notes from IJCAI'01, Seattle, Washington, 6 August 2001, pp. 11-21
-----Original Message-----
From: Horn, Graham
Sent: Wednesday, 3 January 2001 13:06
To: 'John F. Sowa'; pat hayes; West, Matthew MR SSI-GREA-UK;
standard-upper-ontology@majordomo.ieee.org
Cc: Awbrey Jon (E-mail); Fuchs E. Norbert (E-mail); Whitten David
(E-mail)
Subject: SUO: Ontology Structure & Content
John,
. Your mention of lattices at the bottom of your e-mail below
prompted me to return to your "Mathematical Background" at
www.bestweb.net/~sowa/misc/mathw.htm
<http://www.bestweb.net/~sowa/misc/mathw.htm> . I continue to be frustrated
at the apparent lack of comprehension by anyone in the SUO group at what I
have been speaking of as "conceptual dimensionality" and the like.
. Taking your Section 7, I see something close to what I am
addressing.
. Basically, I would group Erdmann and your beverage types in
a different way. The basic attributes I see are:
* temperature,
* constituents:
* minerals,
* plant extracts:
* leaf,
* fruit,
* bean,
* wood,
* bark, and
* root,
* animal extracts:
* meat, and
* milk,
* brewed extracts:
* alcohol, and
* bacilli; and
* "aeration":
* "fizz", and
* froth.
. This provides a far more comprehensive structure, without
even ranging beyond the beverages you mentioned other than to add the
logical additional alternatives. It allows further elaboration, such as from
what species the extracts derived, and what particular extract compositions
are involved. It would thus more readily accommodate such beverages as
cappuccino, cocoa, soy milk, Irish coffee, kava, Bovril, yoghurt and so on,
even flat beer and champagne.
. This involves breaking things up into logical groups that
are independent of each other, and goes into multiple layers of depth.
. The next problem comes when one begins looking at various
perspectives. For example some people may wish to look at stimulant
categories, and wish to group caffeine and taeine together even though one
is a bean extract (coffee, cola, etc), and the other a leaf extract (tea).
(Alternatively, sugar can be either a wood or root extract.) So, does one
break down along the above structure - in which the next level would be the
species of plant from which the stimulant derived, or should one go directly
from the next higher level, and then branch directly into extract types?
Even with this simple example, one can get different structures for such a
reason.
. However, this structural difficulty can be reduced if one is
prepared to allow multiple branchings. For example, the constituents
category above could instead be broken directly down into chemical
categories (eg. some plant extracts are actually minerals, and some
biologically based industrial processes deliberately exploit such an
approach). This would be able to be visualised as the second branching
coming off the page at right angles, because the issue of chemical
composition is conceptually independent of the issue of extraction source.
The two issues are conceptually "orthogonal".
. Please note that this approach does not inherently simplify
the whole universe into a nice easy to read structure. After all, most
beverages are artificial, and we are really breaking out portions of the
ontologies (is this term appropriate here?) of the animal, plant and mineral
kingdoms, etc., at our convenience. Really we should have developed those
more fundamental ontologies, and then chosen to exploit them and their
structures for this beverage classification exercise.
. But I hope this shows some of the power I suggest we can be
using to structure and populate the SUO in a reasonably optimal way.
. Reactions welcomed (Gees, I must be a masochist, going by
the previous reactions I have gotten with this project).
Cheers Graham Horn
Australian Institute of Health and Welfare
================================================
Phone: 02.6244.1094
Fax: 02.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: John F. Sowa [mailto:sowa@bestweb.net]
Sent: Friday, December 22, 2000 9:52 PM
To: pat hayes; West, Matthew MR SSI-GREA-UK;
standard-upper-ontology@majordomo.ieee.org
Subject: RE: SUO: More KIF-ified Ontology Content
Matthew and Pat,
I believe that you have both made some very important points. Rather than
criticizing the ideas I don't fully agree with, I will comment on some that
I like very much.
Then I'll present my preferred way of viewing things, which allows for
multiple viewpoints to coexist (even when they couldn't be simultaneously
asserted without contradiction).
>MW: It is my experience that indeed the things people argue about the most
>are at the highest levels, where as the day to day things people find
easier
>to agree on, because we can point at things to see whether we mean the same
>thing or not. So for example, I doubt if any difficulty we might have on
>agreeing what we meant by "red" would have anything to do with whether we
>saw it as a property, that things could take up, or as a class whose
members
>were states, either individuals, or temporal parts of individuals. I would
>expect the discussion to be around where the border was between red and
>other adjacent colours. So "red" can fit into different frameworks with
>different ways of understanding what being red means.
This is a very clear statement of a position, which I agree with, but which
is more often expressed in a more muddled way, which I don't agree with.
The simpler, more muddled, and to my mind, more dangerous way is the
following:
It is easier for people to agree on the lower-level concepts than on the
higher-level ones. Therefore, the high-level concepts aren't important, and
it doesn't matter if we haven't done a good job on the top levels.
The point where I stop agreeing in such an enthusiastic way is at the next
sentence:
>MW: I would further expect
>to be able to translate between the different frameworks.
Pat has stated my concerns very nicely:
>PH: Obviously the interesting case is where they are referring to
>different concepts. But the difficult case is where A's concept
>cannot even be expressed in B's overall ontological framework, and
>vice versa.
I also like Pat's succinct summary of an important distinction, which I use
as one of the top three in my top-level ontology:
>PH: In the
>continuant/occurent way of thinking, one refers to existence *at a
>time*. This means that the four-dimensional entities simply do not
>exist in this ontology. A continuant is something which continues to
>exist through time and retains its identity through time, and is such
>that (this is the characterising property) all its parts are present
>whenever it is present. Thus a person is a continuant, since if I am
>here then all my parts are here (now), while a race, say, is not a
>continuant - it is in fact an occurrent - since it has parts which
>are temporally distinguished: it has a beginning, a middle and an
>end. The distinction is like that between a person (a continuant) and
>that person's life (an occurrent).
And this is a good summary of the problem:
>PH: Both occurrents and continuants 'map' into 4-d entities in the 4-d
>histories ontology, but there is no principled way there to
>distinguish them. The nearest you can get is to say that continuants
>have isotemporal parts while occurents have isospatial parts, but in
>the 4-d histories ontology one can cut 'parts' any way one chooses,
>including along boundaries which slant in time (ie which are
>'moving', as someone who thinks non-4-dimensionally would say.) And
>the defining criteria for the distinction in the other ontology is
>literally incoherent in the 4-d ontology, since *nothing* is such
>that all its parts are present whenever it is present: the 'whenever'
>here is meaningless in the 4d ontology.
>As evidence for the idea of a continuant in intuitive thinking,
>consider the claim that I am the same person I was 10 years ago. This
>is literally false in the 4-d ontology (it can be expressed there by
>saying that me-now and me-1990 are both slices of the me-history, but
>that raises the question of what distinguishes one history from
>another, since these are also both slices of completely unrelated
>histories) Statements like this seem to depend on the idea of
>something retaining its identity through time, even though its
>properties may change. This 'locus of identity' is what constitutes
>the basic idea of a continuant, I think.
To summarize:
1. Pat says that the 4-D viewpoint allows ways of cutting up the
universe that cannot be mapped into the viewpoint of 3-D plus time.
2. Matthew points out that it is easier to map the 3-D plus time
viewpoint into the 4-D viewpoint.
I agree with both, but only with the caveat that "easier" does not imply
"possible to translate each and every distinction that anyone would ever
want to make."
>>MW: Until recently I used as my baseline a viewpoint that said
that there
>>were individual things like you, my car, classes, and
associations, where
>>an association is a relationship that understands that it lasts
for a period
>>of time. I have moved to a 4D approach, replacing individual
things with
>>spatio-temporal extents, and associations with timeless relations.
However,
>>I know very well how to take something from my old model and
represent it
>>in the new model, and vice-versa.
But Pat points out the problem some distinctions vanish in the 4-D ontology,
and I agree:
>I wonder if you really can do this, if you think hard about it. Ive
>been living within a 4-d ontology for a long time and mapping other
>ontologies to it, and what I find is that many distinctions simply
>vanish, rather than being 'translated'. This is fine with me, of
>course, but it tends to get the other folk a little upset, especially
>when they have written entire books about these distinctions. And in
>the other direction, I find that perfectly reasonable-seeming things
>in my ontology, like 'moving' (sloping in space/time) boundaries,
>simply cannot be admitted into the other ontologies without producing
>unacceptable confusions.
I also agree very strongly with the following:
>PH: I think that what seems to me to be overoptimism about the
> prospects of the SUO among some folk might arise from their failure
> to appreciate that such incompatible conceptualizations can even
> exist, let alone all have their uses.
I agree with Matthew's last sentence, but not the first one:
>MW: I have a simple attitude towards incompatible conceptualisations: they
>mean we do not understand the world around us. The world around us is not
>incompatible with itself.
There may be very good reasons for using incompatible representations even
when a single unified one is understood, as Pat explained with his example
of quantum electrodynamics. I also believe that there are a very large
number of other examples that one could use, as I tried to explain in the
Knowledge Soup chapter of my KR book.
Although I agree with a great deal of what Pat says, I would quibble with
the following:
>PH: the
>world around us IS incompatible with itself, in that we have to use
>two incompatible ways of conceptualizing it in order to fully
>describe it. There is no single underlying theory of the world, and
>some of the best minds who have thought about the matter during the
>last century have concluded that there is no way to produce one.
On the contrary, quantum electrodynamics is a single coherent
conceptualization, and it is possible (at least in principle) to use it to
solve problems without breaking it apart in incompatible ways.
Unfortunately, for most practical problems, QED is so computationally
difficult that different (and incompatible) simplifications are necessary to
get useful results in an acceptable amount of time.
I strongly agree with Pat's point:
>PH: I think that you believe that there is a single, universally
>acceptable, ontology, and that all others can be mapped into it. I
>think this will happen only when science stops.
But I still believe that there is something useful that we can
accomplish even before science stops. My best cut at what
that may be is summarized in the Knowledge Soup chapter of
my KR book. Summarizing even more briefly,
1. An open-ended lattice of theories, which could be extended
arbitrarily far to accommodate any possible way of conceptualizing the world
in any finite set of concepts. A complete lattice would have to be
infinite, but any particular version that might be implemented at any one
time is finite. However, there is no restriction on what anyone might add
to it, given enough time, effort, and ingenuity.
2. Methods for navigating the lattice to find theories that are
approximately true or "good enough" for many problems, and with methods for
revising and extending theories to make them better adapted to solving new
problems.
For more about the lattice of theories, see Section 6 of my paper on
processes and causality:
http://www.bestweb.net/~sowa/ontology/causal.htm
<http://www.bestweb.net/~sowa/ontology/causal.htm>
And by the way, I believe that causality is a very important concept that is
central to a large number of issues in the SUO. But that is another topic.
John Sowa