PatH, (01)
(1) To clarify: (02)
[PH]> > > As you know, Patrick, I don't think this is even
> remotely feasible.
> >
[PC]> >What specifically isn't feasible? Specifically, what cannot be
done?
>
[PH]> Locate a set of basic concepts sufficient that all others can be
> derived from the base set by writing definitions. Now, of course, in
> a sense this is trivial, since any set of concepts can be defined to
> be the 'base'. But I assume that you intend it to be understood in a
> nontrivial way.
>
[PC]
The nontrivial way is to show that the definitions can be constructed
from the constant elements in the foundation ontology, without, in any
one definition, having cross-reference between two or more elements
that are defined in terms of each other, or having to appeal to
subclasses or instances to make the new class distinct from others. (03)
An illustration:
let us assume that the structural elements available in OWL and KIF
are already defined and taken as part of the base. Then I want to
define an "object" (abstract or concrete) as something that has at
least one attribute, and is related to that attribute by the
"hasAttribute" relation; and that objects and attributes are disjoint.
I want to capture the minimal essence of what people think of as
"objects" and "attributes", at least in one common sense - and these
definitions will be used eventually in more specific ways that are of
practical interest. Actually, I want to say more about "objects" (e.g.
they can participate in Events), but this much should suffice to
illustrate the problem. (04)
Then I need to define the object so that all of the constant terms in
its definition are themselves defined by the base ontology, directly or
indirectly. Then to "define" (short for "specify the meaning of
formally, to an adequate level of detail", not to be confused with the
more technical "necessary and sufficient" meaning) an Object, I need to
define an "attribute", and the "hasAttribute" relation. (05)
Problem is, I can't think of any way to define an "attribute"
meaningfully without referring to Objects, and the restrictions that
give some meaning to the "hasAttribute" relation are that subject
(arg1) is an "object" and the object (arg2) is an "attribute". That
is, these three elements can only be defined in relation to each other.
What gives them yet more meaning (to me and I imagine other people) is
in the way they are used, and in the subtypes and instances that we
assign to each category. I conclude that these are primitive concepts
that need to be represented in the foundation ontology. (06)
There may well be cases which, at one time, appear primitive because
they cannot be properly defined without cross-reference, circularity,
or appeal to instances, whereas later, perhaps after some other
concepts have been defined by the base vocabulary, those putatively
'primitive' parts of the base can be seen to be definable using the new
terms, and can be transferred to the supplemented ("practical")
defining vocabulary of useful terms that have been defined relative to
the base. The base will never be frozen, but should, after some
initial expansion, reach a relatively stable state, (like the Cyc
BaseKB is supposed to be). (07)
Does this clarify -- that the base will not be a trivial set of
logical operators, but a set of concepts defined by cross-reference
that are fairly numerous -- or is more needed? (08)
(2) As for:
[PC]> >Perhaps you can provide a set of "challenge" concepts that will
allow
> >us to know when we have succeeded?
>
[PH] > My point is that you will never know this. And you will never
> succeed. And, more really to the point, it will not matter, since
> there is no point it trying to achieve this unattainable goal.
> Suppose for a moment that it were possible and you had in fact done
> it. So what? What use would it be? It would not provide any guarantee (09)
> that it could not all have been done differently, and indeed probably (010)
> is being done differently elsewhere. (011)
There are several points here:
(2.1) [PH] > My point is that you will never know this.
My criterion for success would be when independent domain ontology
developers use some common foundation ontology (an/or its commonly used
extensions) as their conceptual defining vocabulary, and subsequently
transmit information to each other, and when there are relevant
concepts in one domain ontology that are not in the other domain
ontology, they transmit the definitions of those concepts as well.
They then should be able to reach the same inferences from the same
data. That is the goal that I think will provide a level of "semantic
interoperability" that is robust enough for error-intolerant
applications. When users do this, and assert that they have thereby
achieved a level of semantic interoperability on which they can rely
for machines to ***use automated inferencing to make important
decisions without review by a person***, then I will "know this" in
the sense of what I hope to achieve. (012)
Users do such things with carefully crafted software designed to
interact under defined conditions, but not yet (to my knowledge) with
independently developed ontologies. (013)
There are two caveats to this criterion:
(2.1.1) different domain ontologies may have elements other than
those that are transferred for interoperability purposes, and
**additional** inferences in one system may be generated beyond those
in another system. These additional inferences should not be logically
inconsistent with those inferred by that other system, and may or may
not be significant with respect to the purpose of the information
transfer. To be certain of complete consistency, all definitions
beyond the common base that are used in one system would have to be
transferred to the other system, and vice-versa. The effect will be a
common merged ontology. It may not be necessary to transfer all new
elements from one or the other system, it may only be necessary to
transfer a relatively small number that are significant for the purpose
of the information transfer, if such a subset can be identified. If
two groups conduct frequent transfers, it may be simple enough to
identify such a subset for the most common interactions. (014)
(2.1.2) if the two groups actually have different sets of data,
naturally some inferences may be generated by one group using the
additional data and vice-versa. Whether that is significant will
depend on the purpose of the interaction. To be sure that they will
reach exactly the same inferences, again all of the information in one
system will have to be transferred to the other, or they will agree to
work with a common subset of the information that both they have
in-house. (015)
(2.2)
[PH] it will not matter, since there is no point it trying to achieve
this unattainable goal
[PC] I won't argue with your own personal preferences, but I am
assuming that there will be ontology users who will find accurate
interoperability useful (016)
(2.3) [PH] > It would not provide any guarantee
> that it could not all have been done differently, and indeed probably (017)
> is being done differently elsewhere.
Well, it may well be done differently, as for example, by one-to-one
mapping of locally developed ontologies. But at present the defining
vocabulary seems to me to be the only **practical** way I can visualize
to achieve accurate interoperability on a large scale (more than 10
independently developed local ontologies sharing data).
Is it being done elsewhere? We would all, I am sure, be very eager
to learn of cases where people who have applications using
independently developed ontologies are sharing data for
interoperability purposes involving automated inferencing. Such
examples would be enormously helpful to these discussions. (018)
* * * * * * * * * * * * *
Once again, I have no beef with people who are trying to merge or map
ontologies not developed on a common foundation, as I assume they
understand and accept the errors and inaccuracies that will result from
such a method. But for those who want a higher, more accurate level of
interoperability, I am suggesting that the "conceptual defining
vocabulary" tactic provides the most practical path that I have yet
become aware of.
* * * * * * * * * * * * * (019)
>From another note by PatH
>>That is pure stipulation. One might call it more
>>accurately as the Central Dogma of the Semantic
>>Web. (020)
[PH] >Sure, it is a vision rather than a reality; just
> like Patrick's hypothetical Grand Unified
> Ontology. Want to bet which of them will happen
> first?
Hey, I like that idea!! Let's bet. I'll send a separate note on
that topic. (021)
Pat (022)
Patrick Cassidy
CNTR-MITRE
260 Industrial Way West
Eatontown NJ 07724
Eatontown: 732-578-6340
Cell: 908-565-4053
pcassidy@xxxxxxxxx (023)
> -----Original Message-----
> From: Pat Hayes [mailto:phayes@xxxxxxx]
> Sent: Friday, May 04, 2007 4:18 AM
> To: Cassidy, Patrick J.
> Cc: [ontolog-forum]
> Subject: RE: [ontolog-forum] A "common basis"
>
> >PatH,
> > > As you know, Patrick, I don't think this is even
> remotely feasible.
> >
> >What specifically isn't feasible? Specifically, what cannot be
done?
>
> Locate a set of basic concepts sufficient that all others can be
> derived from the base set by writing definitions. Now, of course, in
> a sense this is trivial, since any set of concepts can be defined to
> be the 'base'. But I assume that you intend it to be understood in a
> nontrivial way.
>
> >Finding a set of concepts that will be sufficient to specify the
> >meanings of others?
>
> Yes, precisely that.
>
> > What would it take to prove that it is feasible?
>
> What would it take to show it is feasible to walk to the moon? You
> can try to do it. You will get somewhere. At any point, you can say
> that more work will get more results. All of this is true, and yet
> the goal will always remain elusive. A better question is: how will
> you know when you have succeeded?
>
> >Perhaps you can provide a set of "challenge" concepts that will
allow
> >us to know when we have succeeded?
>
> My point is that you will never know this. And you will never
> succeed. And, more really to the point, it will not matter, since
> there is no point it trying to achieve this unattainable goal.
> Suppose for a moment that it were possible and you had in fact done
> it. So what? What use would it be? It would not provide any guarantee (024)
> that it could not all have been done differently, and indeed probably (025)
> is being done differently elsewhere.
>
> PatH
>
> >Pat
> >
> >Patrick Cassidy
> >CNTR-MITRE
> >260 Industrial Way West
> >Eatontown NJ 07724
> >Eatontown: 732-578-6340
> >Cell: 908-565-4053
> >pcassidy@xxxxxxxxx
> >
> >
> >> -----Original Message-----
> >> From: Pat Hayes [mailto:phayes@xxxxxxx]
> >> Sent: Wednesday, May 02, 2007 6:36 PM
> >> To: Cassidy, Patrick J.
> >> Cc: [ontolog-forum]
> >> Subject: Re: [ontolog-forum] A "common basis"
> >>
> >> >Chris, John, et al:
> >> > I think that a perfectly feasible "common basis" for
> >> >interoperability and integration of multiple ontologies is a
> >> foundation
> >> >ontology structured as a "conceptual defining vocabulary"
> >> that contains
> >> >all the concepts that are necessary and sufficient to specify
the
> >> >meanings of any other more specialized concepts, using
> subclasses,
> >> >relations, functions, and axioms all of whose constant terms
> >> are either
> >> >(1) contained in the foundation ontology or (2) themselves
> >> specifiable
> >> >(recursively, if necessary) by terms in the foundation ontology.
> >>
> >> As you know, Patrick, I don't think this is even remotely
> feasible.
> >> Whats more, it isn't necessary, even if it were feasible. Your
> >> 'foundation ontology', described below, will have
> different points of
> >
> >> view enclosed in 'contexts',and will be freely
> expandable, and will
> >> have ontology-style definitions of concepts in terms of other
> >> concepts, or more generally inferential connections
> between concepts.
> >
> >> This thing, under another name, is already being
> constructed. It is
> >> called the Semantic Web. The ontologies in the SW are your
> >> 'contexts', and the entire SWeb, a world-wide distributed entity
> >> always under construction and never finished, is your 'foundation
> >> ontology'. But the main difference is that, rather than
> being built
> >> by a committee and located in and controlled by a single
> institution,
> >
> >> it is being built by anyone who wants to get involved, is located
> >> nowhere and everywhere, and is owned by nobody.
> >>
> >> Pat
> >>
> >> --
> >>
> ---------------------------------------------------------------------
> >> IHMC (850)434 8903 or (650)494 3973 home
> >> 40 South Alcaniz St. (850)202 4416 office
> >> Pensacola (850)202 4440 fax
> >> FL 32502 (850)291 0667 cell
> >> phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
> >>
> >>
>
>
> --
> ---------------------------------------------------------------------
> IHMC (850)434 8903 or (650)494 3973 home
> 40 South Alcaniz St. (850)202 4416 office
> Pensacola (850)202 4440 fax
> FL 32502 (850)291 0667 cell
> phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
>
> (026)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (027)
|