SUO: Re: CYC Event vs. SUMO Process -- Tomorrow's Sea Battle in a Teacup
o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o
JA = Jon Awbrey
PC = Patrick Cassidy wrote:
PC: I took the time to go through the first dozen of
the messages on the archive list you modestly referred
to -- all messages you sent -- and I couldn't find the
word "event" in any one of them. I am well aware of
the existence of all those wonderful mathematical models
and each one of them will have its place in a Standard
Upper Ontology, and some of them will be very closely
related to Events as I view them, but in creating this
one thread to discuss the spatial location of Events,
I am trying to focus on one very specific point and
I hope that those who reply to this thread will
focus on that point.
I merely gave you a pointer to the thread archive, and what appears first there
is a random topic: "Differential And Riemannian Manifolds", beginning with a few
excerpts from the book of that title by Serge Lang that lays down in quick order the
very standard notion of a manifold -- that serves as a basis for almost all of the
physics that has developed since 1850, not to mention a whole lot of the dynamical
systems theory and engineering that has developed since 1950 -- and continuing with
my try at relating these very standard resources to the very analogous intercomm and
interops problems that press on us here, albeit in a yet unexplored qualitative vein.
http://suo.ieee.org/ontology/msg04769.html
...
http://suo.ieee.org/ontology/msg04782.html
I picked this mathematical text because it is more careful about the
axiomatic aspects of the subject, and so it takes place at a level of
abstraction where application words like "event" and "process" may be
tacit in most contexts, and would be understood as such by everybody
who reads it with an application in mind. But this only goes to show
that what is already considered to be first principles in the fields
that have currently much success dealing with events and processes
in the world simply subsumes under a more "upper, generic, meta"
heading the details that are being made such a big deal of here.
If you have a question about why the theory of manifolds is a potentially
useful resource for our present problems, in particular those involving
change (dynamics, events, processes) or diversity (intercomm and interops
between different reference frames for describing the world), then I think
I am prepared, or at least I will try anyway to say why I think that is so.
I have already been doing that intermittently for a couple of years now.
But it is necessary, when reading in a standard text some peculiar definition
for a familiar word, to ask oneself "Now why did these folks define it that way?",
not to mention taking years and years to come up with that way of talking, and
not just reject it out of hand because it seems to confict with familiar ideas.
And asking the question about events and processes in a way that rules out
all but familiar answers is just another way of rejecting info out of hand.
In statistics, an "event" is a set of occurrences, usually falling
under a given description that happens to be meaningful in a given
context, like "the event of rolling snake eyes". So the meaning of
"event" depends on the meaning of "set" and "occurence", along with
all of the other apparatus of context and description that prevail.
This may seem a bit different from the idea of an event in physics,
having to do with the value of some field that is defined on some
space or spacetime, but it's not all that different, since we are
still talking about sets that fall under formal descriptions,
and the languages of physics and statistics worked out their
terms of endearment, engagement, and interoperation long ago.
So the point is this: For a long time now it has just not made much sense
to ask the question in a non-relative way -- the question about the ontology
of events and processes -- without thinking about the relation to the context
of description that is implied by a sensible answer. A sensible answer can
only talk about events-as-described-in-some-context-of-description. I know
that this does not satify the absolute essence oriented thinker, but before
we can talk about anything more invariant, objective, or universal than
that we have to consider some class of mappings or transformations
among some class of contexts of description.
PC: Your expressions of disdain and skepticism are totally valueless
in trying to create something useful, and I wish you would have
the decency not to junk up a thread which is intended to deal
with a specific issue. There are plenty of other headers
you can use to express you views and demonstrate your
brilliance and superior knowledge.
Look, people who sit down to write axioms for, say,
a Theory Of SpatioTemporal Reality, without knowing
what the average working math or physics person of
the mid 1850's already knew about the subject, are
simply wasting everybody's time and some people's
money, and you have not known disdain until you
find out what anybody who knows anything about
the real "knowledge base" of our times would
think about such a project.
The difference between that level of disdain and my level of concern
is that that I believe there is some reason to build bridges between
formal reason, done by people who really use axioms and models to do
real work, and more approximate, heuristic, informal, and practical
modes of reasoning, which only contradict their own rationale in
trying to adopt an air of formality, solely for the cargo cult
cachet of the word "axiom".
But those bridges are built by starting from the more developed accounts,
already in place and already in use on the contemporary scene -- I think
they call that "Best Practices" -- and extending them in the direction
of more intuitive understanding, not by redoing the 10^2 or 10^3 year
history of their present evolution from ancient topical roots.
PC: In the note below I did not see one specific suggestion about the spatial
location of events. I am not "ignoring what is known" since I am referring
to every one of the half-dozen upper ontologies that have dealt with the
issue in a general way (and a few books on events). On the other hand,
you can't seem to find even one relevant reference to the topic.
Look at the attitude that you constantly express here:
It's like "I'm aware that there's all of this standard
knowledge out there, but I just don't care". This is
the generic attitude that so evidently went into so
many of the contemporary "ontologies" -- in what is
becoming more and more a purely self-referential,
"if it ain't on the web it don't exist", genre of
popular literature. Until there is a genuine desire
to learn what's been going on elsewhere in the world
before they created these intellectual immune systems,
there is little use looking up more references for folks
who never bother to read 'em.
PC: If you want to continue this discussion off-line, I will be happy to oblige.
But if you send a note to the list regarding the CYC/SUMO Event/Process issue,
please make it very specific and to the point. There are plenty of things
I don't know, and I am sure that some of the information you could share
would be valuable in our discussions, if you would only have the decency
and self-control to focus on the issues at hand and not waste people's
time by diverting the topic to things that we don't want to discuss
**at this time**.
Yours is the very sort of defensive reaction to straight information
about what is already standard in the world that will most obstruct
this group from delivering a respected standard on any subject.
It is typical of the response that I have gotten from certain
quarters every time I mention the fact that words like "class",
"formal language", "graph", "information", "program", "set",
"topology", or whatever have several decade or hundred year
histories of technical usages that ought to be taken into
account by Any Body that is seriously entertaining the
notion of dictating standards to others. But sume
people seem to fall somewhere between indifferent
and indignant that anybody might already be
using these words in technical senses that
embody centuries of previous refinement.
Jon Awbrey
o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o
JA: The thing that will prevent any of these characterizations of event and process from
being taken seriously by any research community -- communities whose sense of common
sense is not defined by the maxim "if it's in a textbook it can't be common sense" --
much less viewed as 'the' standard, is that the segments of the SUO community that
generated these characterizations have not taken seriously any of the definitions
of event or process that are already standard in those research communities, say,
engineering, physics, operations research, statistics, systems science, just to
name the ones that I'm acquainted with.
PC: I have read a number of discussion of event and process but
none except those in formal ontologies have attempted to
specify these terms in logical format suitable for
computational purposes, ...
JA: Much to the contrary, the definitions of events and processes in such fields,
all of them being variations on a very small number of themes, all of them
being already axiomatized and build up from basic mathematical concepts
like sets and functions, and typically extended by way of thoroughly
studied mathematical models, are the only ones that are currently
successful in providing critical reflective descriptions of our
most compelling realities, and they form the infrastructure
of all of the truly reliable computational models that
we presently enjoy.
JA: It's true that -- once they leave the nests of their axiomatic bases --
yes, most of these scientific methodolgies currently have a tendency
to issue in quantitative and statistical models that are difficult
to relate to qualitative, logical, and natural language intuitions.
I have personally spent much time arguing with numerical/quantitative
folks that they need to get more lingusitic/qualitative savvy into their
models, but they are not going to do that at the expense of reducing
the descriptive and explanatory adequacy of their current theories.
JA: It must be understood that the mythical number of person years that
have gone into developing systems like Cyc or SUMO are insignificant
in comparsion to the multitude of person years that have gone into
evolving something like the basic notions of a probability space,
a random variable, a stochastic process, a differential manifold,
a 1-parameter Lie group, a transformation semigroup, a program,
and on and on, all of which are highly "generic, meta, upper"
and so on. Nobody in the relevant research communties is
going to abandon any of this just because it's too hard
to re-write it in terms of pre-socratic commonplaces.
PC: ... and relate then to other concepts that will allow us to build
cognitive systems which can support reasoning in a variety of fields.
These **terms** are used in various senses in different communities,
but in the present discussion group we are concerned with creating
logical structures that will be useful in cognitive systems.
Exactly which terms in which language and which jargon words
in which technical community have these logical structures
as their intended meanings is a matter for Natural Language
Processing systems to untangle. If we do a good job, we
should be able to help NLP systems do their job better.
The tasks ar related but distinct.
JA: Promises, promises. These communities have been proving their theorems,
and betting their lives on the results, for hundreds and thousands of
years. They aren't sitting around speculating about the imaginary
powers of next year's vaporware theorem prover.
PC: If you have some specific suggestion with respect to how to represent
the kinds of occurrences in the physical world which people variously
call "events" "processes", "happenings", "occurrences" "situations",
"perdurants", "state changes" or whatever, I would appreciate concrete
suggestions. Your comments above can be interpreted to mean that we
should all take our names off the list and do nothing, since it has
obviously already been done before.
JA: I am typing as fast as I can. Vide:
JA: http://suo.ieee.org/ontology/thrd1.html
JA: My comments should be interpreted to mean that there is a vast amount of work
yet to be done vis a vis creating interfaces between complementary perspectives,
for example, quantitative vs. qualitative models of dynamic processes, but that
progress is not going to come from styles of approach and directions of work
that pretty much exhausted their potential in 500 B.C.
PC: What specific "definitions of event or process that are already standard in those
research communities, say, engineering, physics, operations research, statistics,
systems science" did you have in mind? Can you provide the logical definitions
and axiomatization for our benefit? I for one will be happy to have input from
any constructive source. If in fact they are already standard, can you tell us
what community uses them for what purpose? This thread is intended to discuss
*specifics** of this particular concept class. There are other threads to
discuss general issues.
JA: Look, I can copy excerpts from basic texts until I'm blue in the links,
and I probably will, but it's going to have to occur to somebody that
all of this hard-won knowledge has something to do with the task of
getting actual knowledge in our knowledge bases, instead of thinking
that it would somehow be easier just to scrap it all and start from
scratch. Innovation is possible, but not by ignoring what's known.
o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o