Ali Hashemi wrote:
[AH] >> I am slightly confused. Perhaps you can
elucidate for me what exactly this foundation ontology would assert?
Examples of what a foundation ontology would include can be
found in any of the existing FO’s such as Cyc, DOLCE, SUMO, or (my
favorite) COSMO [http://micra.com/COSMO]. But none of those will serve the
purpose of the common FO that could be developed by a broad consortium of developers
and users, because each of those has a structure that makes representation
choices different from the others, with no provision for translation into other
views. The common FO would have enough of the most primitive conceptual
elements to allow views represented in one ontology to be translated into the
form of any different ontology, by means of bridging axioms, or if necessary
procedural code. In addition, to be widely acceptable, the common FO would
need utilities to make it easier to use than existing FO’s, and in
particular a good natural language interface.
The most important function of the FO is to provide all of the
basic elements to allow anyone to create a domain application ontology in which
all of the domain ontology elements are created as combinations of the
pre-existing elements in the FO. This will allow automatic interpretation of
any domain concept developed from the preexisting ontology elements, whose
semantic are well specified and agreed on by all users. That automatic
interpretation is what I call ‘semantic interoperability’. I
call the FO the ‘conceptual defining vocabulary’ because it
provides a means to specify the logical structure of any domain element (‘defining’
the domain elements, in a loose dictionary sense).
The apparently difficult conceptual barrier to overcome in visualizing
this process is to accept that there can be agreement on such a basic
vocabulary. The way to overcome this barrier is to realize that **anything**
that **anyone** thinks is necessary to specify the logical structure of their
domain ontology elements can be included in that FO. What may be even more
surprising is that very few of the elements needed for such a basic vocabulary
are actually logically incompatible with each other – they merely
represent different ways of viewing the same entities. This is a conclusion I
have reached after 15 years of paying attention to debates on that issue. An
example of the problem is the issue of whether 3D and 4D perspectives can be
translated into each other. Both Pat Hayes and I have presented examples of
bridging axioms that can translate assertions formulated in either view into
assertions formulated in the other view. Other cases of supposed logical
incompatibility can also be resolved by bridging axioms. The elements of the
FO itself will be logically consistent. In cases where some domains need to
represent logically inconsistent concepts, they will be represented in
extensions to the FO. But in each case, the elements of those extensions can
be logically specified using only the elements of the FO. What this means is
that not only *we* can recognize that these theories are incompatible,
but the *machine* can also recognize that the theories are incompatible,
**while still understanding what the theories are asserting**!! What happens
with incompatible theories is that they are not all asserted as true elements
of a single ontology, but as separate theories, whose axioms are not asserted
as part of the FO itself. The segregation of theories form each other is a
common tactic – in Cyc it is accomplished by microtheories, and by
contexts in other ontologies.
[AH] > Yet, if you are asserting that a physicist,
biologist, anthropologist, lawyer and psychologist would somehow agree on the
same fundamental ontology, I have difficulty seeing such a proposal coming to
I am asserting that, and I believe I have very good reason. As an example, let
us say that we know that there are incompatible theories of space and time –
Newtonian and Einsteinian. Their *models* are logically incompatible.
Each of these is therefore represented as a different *theory* of space and
time. But both of those theories can be adequately represented using the same
basic inventory of primitive ontology elements that are themselves logically
compatible and all maintained in the FO. So, not only do *we* know that
there are two ways to calculate the trajectory of an object through space, but
the *computer* can also know that. And the computer can also know that
each of these models yields an *approximation* to the actual behavior of
objects in space, and that the Einsteinian approximation is more accurate but
harder to calculate than the Newtonian approximation, and that the Newtonian
approximation is good enough for use in specific circumstances and that
Einstein may be needed in other circumstances. Representing all of that basic
knowledge of how to deal with different circumstances is the task of the FO.
It will be complex, and it will need a large community to demonstrate that it
is feasible, by creating applications that use it to good effect.
Of course building applications to demonstrate the utility of an
FO is not simple. If it were it would have been done long ago. That’s
why it needs a scientific community of substantial size to develop and test and
evolve the FO and its essential utilities, in particular the NL interface; and
then to create publicly available applications that demonstrate its use.
Meanwhile there are plenty of more modest applications that do
not require an FO at all, and do not need to interoperate with other applications,
so people will develop their local ontologies and hopefully use them to good
effect. If they want to interoperate with more than a few other ontologies,
however, an FO is the most effective way to do so.
There is powerful potential in a community that adopts a common
FO, and I am suggesting that that potential be properly explored.
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Ali Hashemi
Sent: Tuesday, May 12, 2009 6:32 AM
Subject: Re: [ontolog-forum] Ontology Project Organization:
I am slightly confused. Perhaps you can elucidate for me what exactly this
foundation ontology would assert?
Are we talking here of say, properties of unary, binary, ternary etc. relations
devoid of any particular conceptualization? If so, then yes, we may reach some
sort of common foundation.
Yet, if you are asserting that a physicist, biologist, anthropologist, lawyer
and psychologist would somehow agree on the same fundamental ontology, I have
difficulty seeing such a proposal coming to practicable fruition.
Note in my response, I don't disagree that there need be an overriding
framework to make sense of the sundry ontologies now available. However, I
suppose I disagree with you in that the solution is a foundation ontology
(unless of course you meant one for the abstract properties of relations and
I appreciate your description of the current state of affairs, but before I
write more, I would be very grateful if you would provide an example(s) of something(s)
the foundation ontology would include.
On Mon, May 11, 2009 at 7:52 PM, Patrick Cassidy <pat@xxxxxxxxx> wrote:
Re: > Just
want to reiterate one basic point -- the idea of a single foundation
ontology is fundamentally, and fatally flawed. I think you've agreed with this
suggestion in the past as well, but I see the term being used here in the
No, one single foundation ontology
is **absolutely essential** in order to support *accurate* semantic
interoperability. Without a common standard of meaning, ambiguity and
misinterpretation are inevitable. There is plenty of room for
multiple ontologies of particular domains, including basic things like
time. But without a single common foundation ontology that can
serve to *translate* among the other ontologies, there is no hope for
accurate interchange of information. There is no flaw here, a
common foundation ontology is the whole point, to allow independent development
of multiple domain ontologies that can interoperate.
The current situation indeed
demonstrates that people will generate their own local and different
ontologies, **but** that is occurring precisely because there is no widely
accepted common FO. If there were one, and if it were usable by the
public, and if anyone wanted to interoperate with other systems, it would be
madness to go off and develop an incompatible ontology, because it would be
totally unnecessary, costly, and prone to repeat errors that were made and
corrected in the past. The point of the common FO is to contain all of
the basic elements that anyone would need to represent what they want to
represent, in the way they want to represent it. All they need to do is
define their domain elements by use of the common FO.
It is important that we not
confuse what happens in the absence of any widely accepted common FO from what
would happen if there were one. The point of the FO development project
is to create an active user community so that there will be **at least one**
common FO that people can use. Then, if anyone wants to interoperate with
others, there would be a very powerful incentive to use the most widely
used existing system.
Yes, people develop their own
ontologies (and I do too!) because they have no reason to want to interoperate
with any existing system. There are too few users, and no public
applications. When, inevitably, some large community does arise that uses a
common FO, the whole situation will change totally. With other real
applications that one can interoperate with, and an FO capable of representing
anything you want to represent, who in their right mind would go off and
develop a different ontology anyway? Only those who do not care about
interoperability, which is almost everyone who has developed their own ontology
thus far. If there were already a widely used FO, with public
applications and an NL interface, I would use that instead of investigating
alternatives. But there aren’t any.
Please do *not* *not*
*not* confuse the requirements for developing a useful local application
with the requirements for semantic interoperability. They are totally
different. So far, there is no broad user community for any FO, and the
inevitable result is that people develop their local ontologies that are
congenial to them, having little incentive to use anyone else’s and
powerful disincentive (they are hard to learn and to use).
The point of an FO project is
to *create* a large user community so that people who want to
interoperate with others will have a community that they can participate
in. If local incompatible ontologies are adequate, use them.
If you want broad semantic interoperability , you must find a large community
that uses a common FO.
Perhaps the most important
thing to remember is that the potential user community is massively larger than
the tiny band of people who develop their own ontologies with alternative
representations that differ from each other. Most users will be
database-driven application developers who don’t care a hoot about
ontology, but would be thankful for a convenient way to develop their local
data models in a way that permits accurate semantic interoperability with other
DB developers. When an FO is widely used to support interoperability
among applications, the earlier history of ontologists going off and developing
separate systems, any one of which would do the same job, but which are
difficult and impractical to translate post-hoc, will be looked on as a quaint
distant episode from the past.
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
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 (01)