On Jan 6, 2009, at 10:58 AM, Patrick Cassidy wrote: Ed,
The point you miss is that the foundation ontology does not need to take
a stance on **any** theories that are contradictory, it merely needs to
provide the conceptual vocabulary with which to **describe** theories people
build.
But that is not an ontology: it is a meta-theory of alternative ontologies. This might indeed be a worthy research effort, but its important not to confuse it with actual ontology-writing. Alternative theories can be represented in extensions to the
foundation ontology
That is extremely unlikely, for two reasons. First, no such foundation is likely to exist; second, even if it did, the extensions to it would be mutually inconsistent with one another, so would not all together form a coherent single ontology. (or in John Sowa's terms, as a 'lattice of theories').
In that way, the similarities and differences among theories can be
logically specified, and where there are consistent though syntactically
different representations, they can be translated into each other. No
ontology can make people agree on inconsistent models, but having the
logical vocabulary to properly describe the differences will allow people
(and machines) to recognize where the differences are, how the differences
affect practical issues, and whether there is data that can refute any of
the proposed models.
Again, you seem here to be talking about a comparative meta-theory of alternative ontologies, a kind of uber-meta-ontology which talks about ontologies rather than about the world that those ontologies describe. This is an interesting-sounding idea, but I have no idea how to begin approaching it.
My confidence comes from 15 years of looking at all of the examples
(there aren't many) of "inconsistent" ontological representations, and
realizing that they fall into one of two different categories:
(1) different models.
Consider Newtonian and Einsteinian space.
For simplicity let us say per Newton "space is flat(Euclidian)"
And per Einstein "space is curved"
But this begs the question. Calling Newtonian space "flat" is only meaningful from Einstein's viewpoint. In Newton's ontology, space isn't either flat or curved. The idea of its having these attributes is (literally) meaningless.
The foundation ontology has the meanings of 'space', 'flat(Euclidian)'
and 'curved',
And the attribute relation 'is' encoded. It can describe the two
theories.
Then its already Einstein's theory.
The two theories themselves are contained in a "physics" extension
to the
Foundation ontology. More detail can be added as needed, using a
single
logically consistent representation of the basic concepts (i.e.
the foundation
ontology)
Maybe you could share with us a sketch of this logically consistent basic physics which has Newton and Einstein as extensions? Good luck. (Seriously: isn't it surely just obvious that this is pure fantasy?)
I believe that this is in essence similar to the way that Andreas Tolk
describes
His approach to alternative models. I view the foundation ontology
as the
set of concept representations that do *not* depend on models for
which
there are disputed alternatives, and I believe that that set of
logically consistent
ontology elements is large enough to serve as the conceptual
vocabulary with
which all of the alternative models can be described.
This seems vanishingly unlikely. This first requirement forces this common theory to be as small as possible, but you believe that it will be as large as any theory every conceived by the human mind. One point
that Andreas made,
if I interpret him properly, is that it often occurs that what
people view
as different "models" are actually different *views* that are
logically
consistent, but contain representations of different aspects (e.g.
attributes)
of the same entity.
This may be true in many cases, but it is most definitely not true in the hard cases that cause the most grief, such as the notion of a 'continuant' which underlies the BOF and DOLCE ontologies. In a 4-d ontology, being a continuant isn't a missing attribute: its impossible. So here is DOLCE which says that people are continuants, and here is EPISTLE which says they are 4-d. So what am I, in this combined ontology? I can't be what DOLCE says I am, in EPISTLE, since in EPISTLE it is fairly easy to show that none of those things can possibly exist. The DOLCE view is that EPISTLE has me mis-classified as an occurrent. The EPISTLE view is that everything in space and time - every physical thing - has the attributes that DOLCE restricts to events.
The key point is that ontologies do not only assert attributes that things have. They also make assertions about attributes that things cannot have. If you believe in continuants, then continuants cannot have temporal parts. It is inconsistent (or an error of some kind) to even talk about such things. If you think in the EPISTLE way, then of course anything that lasts a time has temporal parts: that is a necessary truth in EPISTLE. What kind of basic theory can reconcile these differences of opinion? In such cases, both *views* could be
represented in
a consistent ontology, because they can be translated into each
other, using the
basic concept representations.
They cannot be translated into one another at all, with or without any basic representations. They disagree about the fundamental structure of reality. They are in metaphysical opposition.
(2) terminology disputes
The function of a foundation ontology is not to force anyone to accept
another
Person's use of a particular term, but to create logical
specifications of terms
So that similarities and differences in term usage can be accurately
represented.
In one example of this problem, a very competent ontologist described
a session in which
there was a discussion of the meaning of being "inside" a structure;
is the paint on
the wall inside or is it part of the structure? When one views the
foundation ontology
as a defining vocabulary, the answer is quite obvious - to say the
paint
is "inside" is one meaning of "inside" (insideIncludingPaint), and if
one
wants to say the paint is part of the structure, and no part of the
structure is "inside",
then one creates a different relation (e.g. insideNotIncludingPaint or
insideNotPartOfStructure)). If both relations are useful, then both
need to be included in one's ontology. They are not inconsistent
unless one insists on giving relations with two different meanings the
same label.
But this answer is too glib. First, the key point of that anecdote was that these people had been using the same English word all their lives, and would both have said that they were in perfect agreement on its meaning; and the ontological difference that was revealed was invisible in any dictionary definition. (The moral being that one cannot trust either dictionaries or naive introspective intuition when comparing ontologies.) So forcing them to make a distinction that neither of them apparently found useful isn't likely to lead to a successful tool. But in any case, it seems obvious that these two notions of 'inside' are very closely related, to the point where most people for most purposes wouldn't want to distinguish them. [1] What, in your vision, records this important semantic fact? If we just coin two names for them, these two are no more logically related than sugar and salt. Third, this simple strategy is bound to explode in your face immediately. Cyc has already tried this, which is why it needs (and has) something like 12 different senses of 'cover', and its not hard to find examples which will split any of those into sub-cases when more can be said about the things covered and covering. (For example, if a balding man grown his sparse hair extra long and combs it to cover his baldness, in which sense is the bald scalp covered with hair?) The resulting explosion of closely related but logically distinct relations and concepts rapidly becomes overwhelming even for human thinkers, let alone machine reasoners.
[note 1] In fact, my analysis of that story was that they were using the same notion of 'in', but different notions of 'room' or 'office'. But again, this analysis could be disputed, so maybe we would finish up with splitting the 'in' relation and also the 'habitable space' concept, giving 2.2=4 alternative readings of the English sentence. These multiplications expand more rapidly than one can keep track of, with no sign that they will eventually stop; and indeed, no logical or engineering technique available for preventing them expanding beyond any known limit.
. . .or combinations of both.
When looking over Pat Hayes' catalog of time theories, there were
indeed different
"theories" that depended in different models of time, but they could
all be
described using the same basic conceptual vocabulary.
No, they couldn't. The notions of 'point' and 'interval' were all dependent on the particular ontology. Thats why I called it a catalog of theories rather than a theory. In the
foundation ontology
one can have open intervals, closed intervals, and semi-open
intervals. They
are all different things, and confusion arises only if one tries to
give two
different concepts the same label.
You completely miss the point. You are assuming that the intervals being described are the usual intervals on the real line, but that completely begs the question. Some of those temporal theories don't even have the real line as a possible model. In most of the ontologies, there was no distinction between open and closed intervals. In some of them, an interval could be both open and closed. In others, these notions could be defined but did not have their usual mathematical meaning because the underlying order structure was different. There is no single model of temporal ordering. Some ontologies allow time branching, others don't. Some allow alternative futures to converge, others don't. Some allow closed intervals to meet, others don't. The Allen interval algebra works in many of them but not all of them. The only ontology common to all of them is the weakest one possible, which amounts to saying that points are (at least) partially ordered and intervals have points at their ends; and this alone isn't hardly enough to do any useful temporal reasoning at all, certainly not enough for temporal planning. Likewise, one may have two
different
functions called generically "concatenation"; to concatenate two
"glass intervals" and have the resulting interval continuous is a
*different*
Function from concatenating those two intervals and having the
resulting
Interval missing a point in the middle. No problem occurs as long as
One doesn't try to give the same label to two different functions.
Its not that simple. In many of the ontologies, there isn't any glass continuum; in fact, in the conventional real line theory one can prove that there are no glass intervals. Both of these common cases are consistent with, and can benefit from,
using a logically consistent set of ontological representations of the
primitive concepts that can accurately describe the alternatives.
Plain flat wrong. Provably wrong, in fact, in this case. There is no such set of primitive concepts for these ontologies. (Think about it for a second: the only concepts in most of them were those of timepoint and time-interval, and in all but the simplest, these are in fact interdefinable, so one can make do with one. What more "primitive" concepts can one reduce this set to?)
Perhaps you can suggest a very specific case where there is a logical
inconsistency
that *cannot* be described using a single logically consistent set of
concepts?
The time catalog will do as a first example. It documents many of the inconsistencies in the text.
If you accept this description of the structure and function of a common
foundation ontology, there still remain two questions:
(1) how many basic concepts are there that can be agreed on, once one
agrees
That all **logically consistent** views (translatable into each
other) can be
Represented in the foundation ontology itself, or in some internally
consistent
Extension module?
(2) Will the inventory of logically consistent basic concepts in fact be
large
Enough to ontologically describe any data structure in any
application?
If not, how frequently will the foundation ontology need
supplementation?
I don't know the answer to these questions, and they can only be answered
by a serious attempt to build such a foundation ontology, and test it
systematically for its adequacy as a conceptual defining vocabulary.
Do you know of a proof that its impossible to build a Lofstrom launch loop? ( http://en.wikipedia.org/wiki/Launch_loop) The only way to find out by a serious attempt to build one. Let's all get down to it! I
believe that it is important to answer those questions, because if we can
identify a necessary and sufficient inventory of primitive concepts that
will give us a powerful tool to enable accurate semantic interoperability by
using the foundation ontology to translate information from one viewpoint to
another.
Very doubtful. First, this would be true only if everyone agreed to use the same foundation ontology, and there is not the slightest shred of evidence to suggest that people or organizations will be willing to do this. The army currently has thirteen incompatible knowledge models in use, all of them with a large user community and a huge associated investment in training and software development, everyone agreeing this a a huge problem, but no forseeable prospect of anyone agreeing to adopt any of the others. And this is in a controlled military environment where the grunts at the bottom have to obey orders. Second, its simply not clear that the existence of a single common ontology does in fact make interoperation easier, although this is often asserted as though it was self-evidently true. Until a convincing demonstration is forthcoming, I think the jury should remain out on this. There are many reasons why being forced to use a highly developed ontology might make knowledge interoperation harder to achieve.
Do you have no curiosity at all about whether such an inventory of
primitive concepts exist?
I did, briefly, when I was a graduate student. I quickly convinced myself that it does not, partly by watching Roger Schank try to do it. Every attempt I have seen to create one has been a dismal failure. The idea is now firmly placed in the realm of recurrently occurring mad ideas, like perpetual motion and cold fusion. Do you think that answering this question is
actually valueless?
I think that for all practical purposes the question is already answered, in the negative. In 30 years I have never seen a good argument in favor of it. You havn't offered any, only your repeated insistence that it must be possible and the observation that if it were, things would be a lot better. But of course the same is true of perpetual motion and cold fusion (not to mention a Lofstrom loop.)
PatH Pat Patrick Cassidy MICRA, Inc. 908-561-3416 cell: 908-565-4053 cassidy@xxxxxxxxx-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-
bounces@xxxxxxxxxxxxxxxx] On Behalf Of Ed Barkmeyer
Sent: Tuesday, January 06, 2009 1:00 PM
To: Patrick Cassidy
Cc: '[ontolog-forum] '
Subject: Re: [ontolog-forum] Next steps in using ontologies as
standards
Patrick Cassidy wrote:
Your gut skepticism is not surprising, but I tried hard to find
some
objective criticism of the proposal in your response, and couldn't.
Interesting. I need to write more clearly. Have you looked at the
collection of biological ontologies in Mark Musen's Bioportal? How
many
different models of human anatomy did you see? Why is there not one
consistent model? And what is the secret formula for removing that
obstacle?
Ask Steve Hobbs about the effort to develop OWLTime, which is a very
"small" fundamental ontology in your vast undertaking. They had to
choose among three competing models of time, after giving up on
usefully
reconciling them. Pat Hayes' paper documents the fundamentally
different axiomatic approaches. According to you, a well-funded
community of great minds will find one model they can all agree on.
What makes you think so?
My gut skepticism is based on published experience. My apologies for
failing to mention more than the biological work and the conflicting
standard upper ontologies. With a bit of effort, I can find a few
others. But my mathematical training tells me one counterexample
disproves a theorem.
What evidence is your unbridled optimism based on?
-Ed
--
Edward J. Barkmeyer Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263 Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263 FAX: +1 301-975-4694
"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority."
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
_________________________________________________________________ Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/ Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/ Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx Shared Files: http://ontolog.cim3.net/file/ Community Wiki: http://ontolog.cim3.net/wiki/ To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
------------------------------------------------------------ IHMC (850)434 8903 or (650)494 3973 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32502 (850)291 0667 mobile
|