| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Graham, et al.
It seems to me (and Nicola Guarino) that when building an ontology one is
(theoretically) faced with a number of major ontological choices, that have
a significant impact upon the overall structure - the architecture. I say
theoretically above, because it seems to me that we only become aware of the
choices and their inter-relationships in retrospect - after we have had some
experience of building ontologies. So in building second (at least)
generation ontologies we have the opportunity to try and articulate the
choices that significantly impact the architecture - and understand what
motivates them and how they influence one another. A coherent ontology, for
me at least, is one where these choices have been made in a sensible and
consistent way. One can do this by intuition, but it is easier - especially
when working with a number of people, to make the options explicit.
Nicola and I have been discussing how this may be done. I have made a first
attempt at this looking at some 'choices' that impact on the 3-D and 4-D
debate. The two choices I focus on are:
· Perdurantism versus Endurantism
· Presentism versus Eternalism
Other choices that I suspect need looking at are things like:
· Absolute versus Relative Space, Time and Space-Time
· Modally extended versus unextended individuals
· Materialism and Non-Materialism
· Extensionalism versus Non-Extensionalism - I - Universals
· Extensionalism versus Non-Extensionalism - II - Particulars
· Topology of time - branching or linear.
I found that the explanation of the choices was longer than I wished, but
having been exposed to the range of people interested in the SUO it seemed
sensible to try and make the explanation as clear as possible - so that we
could focus on the choice rather than argue about what the choices
'actually' were.
I would be interested in any comments - bear in mind that this is a rough
draft hastily put together. The kind of comments that I think are most
useful would be structural - e.g. the structural implications of the
continuant and occurrent categories being disjoint.
I have not tried to correlate the choices to Ian Niles's merged ontology as
I think these are a pre-cursor to it. However, one point may be worth
noting. I have made a reference to continuant and occurrent in the text. I
have used what is I understand the usual meaning (at least in the
books/articles that cross my desk). Ian Niles's merged ontology seems to
have taken a different meaning to these.
Regards
Chris
-----Original Message-----
From: owner-standard-upper-ontology@ieee.org
[mailto:owner-standard-upper-ontology@ieee.org]On Behalf Of Horn, Graham
Sent: 02 February 2001 11:56
To: 'mail@ChrisPartridge.net'; 'Schoening, James R CECOM DCSC4I';
Standard-Upper-Ontology (E-mail)
Subject: SUO: RE: RE: #8 "No better than Cyc"
Thanks Chris,
. Are you able to elaborate on the meaning of
"coherent", and possibly give an example or two?
. Further comments interspersed below, prefaced "GH> ".
Cheers Graham Horn
National Data Standards Unit
Australian Institute of Health and Welfare
================================================
Phone: +61.2.6244.1094
Fax: +61.2.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: Chris Partridge [mailto:chris_partridge@csi.com]
Sent: Friday, February 02, 2001 9:24 PM
To: Horn, Graham; 'Schoening, James R CECOM DCSC4I';
Standard-Upper-Ontology (E-mail)
Subject: RE: RE: #8 "No better than Cyc"
Jim (Graham),
One of Cyc's important shortcoming (in my view) is that it is not
(philosophically - whatever that means) particularly coherent. This is
common in the early attempts to build systems. This is also a criticism of a
number of people of the so-called 'merged ontology'.
GH> Is there any reason why, with a bit of guidance, Ian can't adapt the
merged version to overcome this criticism? Anything that smacks of internal
inconsistencies is naturally undesirable.
It would seem to me that the SUO provides an opportunity to do this better.
I think it would be helpful to list the architectural options for an upper
ontology - so that we could make informed decisions about them. This might
be a better basis for proceeding in this area than a list of Cyc's failings
- that is not to say such a list would not be useful, as would want to learn
from their 'mistakes'.
GH> I agree. We should do both. (The way you worded this, it isn't
evident you see benefits for both until one gets to the last half of the
last sentence above - till then it looked like you felt it was an "either
or" situation.).
Regards
Chris
-----Original Message-----
From: owner-standard-upper-ontology@ieee.org
[mailto:owner-standard-upper-ontology@ieee.org]On Behalf Of Horn, Graham
Sent: 02 February 2001 04:20
To: 'Schoening, James R CECOM DCSC4I'; Standard-Upper-Ontology (E-mail)
Subject: SUO: RE: #8 "No better than Cyc"
Jim,
. This raises the issue of the nature of Cyc's shortcomings.
. Can anyone list and/or explain what they are, preferably
including the practical implications such as what they prevent the ontology
from covering?
. This would be very helpful for some of us to appreciate the
nature of practical concerns we need to worry about.
Cheers Graham Horn
National Data Standards Unit
Australian Institute of Health and Welfare
================================================
Phone: +61.2.6244.1094
Fax: +61.2.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: Schoening, James R CECOM DCSC4I
[mailto:James.Schoening@mail1.monmouth.army.mil]
Sent: Friday, February 02, 2001 9:49 AM
To: Standard-Upper-Ontology (E-mail)
Subject: SUO: #8 "No better than Cyc"
Comment #8: Bob Spillers [mailto:skydog@pacbell.net] posted a comment (full
text below) that basically says the Merged Ontology, if done well, will
"produce an ontology that is different [than Cyc] , but about as useful as
Cyc," which he says is "not what I would prefer as an upper level ontology."
In case I have not synopsized this accurately, his full comment is below.
Comment please.
Jim Schoening
Sent: Wednesday, January 31, 2001 10:50 PM
To: Jim Schoening; standard-upper-ontology@ieee.org
<mailto:standard-upper-ontology@ieee.org>
Subject: SUO: Vote on Merged Ontology
I vote NO.
Before commenting on the proposed ontology, I think it worth while to point
out that, in my view, this vote is unnecessary and divisive. I think it is
poor judgment on the part of the chair to call for votes on his own
initiative and without discussion of the wisdom and timing of such a vote.
It puts people in the difficult position of wanting to be supportive of work
being done but having to say the work is not a satisfactory basis for a long
term effort.
This situation could have been avoided by first determining that a vote was
desired by the participants and before calling for a vote, clearly stating
how the vote will be counted. I mean basic things like how many people are
eligible to vote (at this particular time), who the eligible voters are
(name), how the votes will be counted (e.g. does ABSTAIN count as NO), is a
super majority (60%, 75%, etc.) required and how will a percent YES (or NO)
be determined - as a percent of those voting or as a percent of those
eligible to vote - is a quorum required, if so how large a quorum.
Standards organizations have carefully defined rules about voting. The rules
vary among organizations and situations, but whatever they may be they are
clearly stated. In many cases the rule is 75% of those eligible to vote and
75% of those actually voting must vote YES (ABSTAIN does not count as YES).
This may or may not be appropriate for this group, but some clearly stated
procedure is required.
Since there have been several statements that the IEEE is silent about
voting procedures in this type of group, as a separate matter I will propose
a set of voting rules for the group's consideration, comments, revision,
etc. If the group can come to a general agreement on a particular procedure,
then I will move that they be formally adopted.
I agree with the technical criticisms offered by Pat Hayes, John Sowa,
Nicola Guarino and others about the specifics of this particular ontology. I
also believe the technique of alignment / merger will not produce an
ontology that will be widely accepted. This is in large measure the approach
taken by Cyc and they have done it reasonably well. A similar approach, if
done well, is likely to produce an ontology that is different , but about as
useful as Cyc. If it is done well it will be costly and it seems to me this
would be a waste of considerable resource.
I am not arguing that Cyc be accepted as "the" standard upper level ontology
- Cyc already is the "standard" to the extent there is one. Doug Lenat
(Cyc's CEO) offered his co-operation in one of the first messages posted to
this list (I couldn't find this message in the archives). The Cyc upper
level is not what I would prefer as an upper level ontology. It is what I
recommend to clients that need one -- for commercial purposes, it really is
the only game in town. It also has the considerable advantage of working
with Cyc's lower level ontologies.
Doug is willing to convert to a "Standard" if a reasonable one is developed.
I will quote a paragraph from a note I submitted to the list in May that I
am also unable to find in the archives.
In previous discussions with Doug, he has been willing to adopt a
standard reference ontology should one emerge. The caveats are (1) that it
is well constructed (2) that it enables Cyc to do things that it currently
cannot or cannot without great difficulty (3) that the expense of adopting a
new upper level be reasonable (4) and that the discernible advantages to Cyc
from (1),(2) & (3) above are not marginal. (-- Doug, I hope this is an
accurate representation of your views.) I believe this is reasonable and
also sets a default as to what is worth doing. Another upper level ontology
about as good as Cyc or somewhat better would not be
worth the scale of investment that I think will be required.
The approach that I believe will produce a satisfactory result is along the
lines of Nicola's suggestion that a small group of highly competent people
meet to hammer out a basic philosophical / logical approach as a foundation
for a computational artefact (an upper level ontology) that will be broadly
useful and widely accepted. After (and only after) such an agreement can be
reached, the concepts that have already been examined, described,
defined/partially defined, axiomized, etc. can be properly placed in the new
ontology.
This is not a merger of existing ontologies. It is the creation of a new
ontology -- and where appropriate, using some of the content of existing
ontologies in a principled way. If the concept(s) and its relation(s) have
been considered for use in one ontology, the intellectual capital this
represents is a huge advantage over starting from scratch, even if the
concept is used differently or not at all.
A criticism of this method is that after 2500 years of disputation, it may
seem overly optimistic to expect agreement on basic philosophical / logical
issues - currently there are strong disagreements among well known, vividly
articulate personalities. My response is to avoid attempting a resolution of
basic philosophical issues by asking for agreement where possible; when
agreement is not possible, compromise on the implementation of these issues
in a way that doesn't do violence to deeply held beliefs. One is asking for
agreement on the design of a computational artefact - not realism vs
nominalism, etc. Newtonian vs relativistic physics might be useful as an
analogy. Although one may know that the universe operates differently in
fundamental ways, Newtonian physics solves the vast majority of problems
encountered in anything like ordinary experience (including space travel -
at least the travel done to date).
A first attempt at this approach was made at the Heidelberg Conference. This
was a private, invitation only conference held at the European Media Lab
located at the Villa Bosch in Heidelberg with generous support from the
Klaus Tschira Foundation - a private German foundation that also supports
the European Media Lab. The theory group was composed of people who are
competent and well known in the field. Among them are some who have
contributed to this list. The participants were chosen both for their
competence and because they are well known to have strongly held and
differing views. The idea being that agreement among them would have a
chance of broad acceptance.
It is an understatement to say that ideas were challenged -- so was
intellectual ability and parentage (among other things), people stormed in
and out of meetings, furniture was broken... After about ten straight days
(of 10 to 12 hour days) an agreement was reached concerning many of the high
level (and most difficult) issues. Much of what was agreed upon were what
factors should be used (or not used) in a factored combinatorial structure
that results in a lattice. This structure is somewhat similar to Formal
Concept Analysis that was mentioned by someone in an earlier note to this
list - however the discussion did not centre on category theory vs set
theory, etc.
This work needs to be fleshed out and it takes a serious effort that cannot
be done on a volunteer basis. I am hopeful that funds will be available soon
for this purpose. If so, it will be only a beginning.
Creating a standard upper level ontology - a reference ontology - will be
costly, connecting it up to natural language (common English for starters)
will be very costly. It is naive to think otherwise. It is also naive to
think that the proposed merged ontology can serve as a basis for a standard.
However, it is important intellectual capital that should be encouraged and
preserved.
Bob
Ontological_Architecture_01-02-02.doc