Hi: (01)
>From my experience, one way to approach this aligning/integration of
many ontologies/processes is to integrate them to a common ontology that
"encompasses" them all, which requires an "all-inclusive" perspective of
the common ontology. That's what I'm suggesting with the general
ontology approach I've been describing in this forum. (02)
If you don't take this common-ontology approach, they you're going to
have to build a mapping from each unique functional ontology to each
other functional ontology, leading to all the possible permutations that
might come from this. Building a single interface to a common ground or
backplane is better than building one for each possible use, or so
industries/groups that do a lot of integrating have found. (03)
Roy (04)
P.S. When this change from a listserv from "ONTAC-WG General
Discussion" to a mass mailing with so many email addresses? (05)
-----Original Message-----
From: Cory Casanave [mailto:cory-c@xxxxxxxxxxxxxxxxxxxxxxx]
Sent: Wednesday, November 30, 2005 12:39 PM
To: 'Adam Pease'; 'Cassidy, Patrick J.'; 'Anne Cregan'; 'Antoinette
Arsic'; 'Arun Majumdar'; 'Barry Smith'; 'Brand Niemann'; 'Charles
Turnitsa'; 'Christopher Spottiswoode'; 'Dagobert Soergel'; 'David Eddy';
'Eric Peterson'; 'Gary Berg-Cross'; 'Harold Frisch'; 'I.N. Sarkar '; 'J.
P. Morgenthal'; 'James Schoening'; 'Jeffrey A. Schiffel'; 'John F.
Sowa'; 'John Thompson'; 'John Young'; 'Leo J. Obrst'; 'Lowell Vizenor ';
'Matthew R. West'; 'Michael Denny'; 'Michael Gruninger'; 'Nicolas
Rouquette'; 'Olivier Bodenreider'; 'Patrick Cassidy'; 'Pawel Garbacz';
'Peter Yim'; 'Richard Lee '; 'Richard Murphy'; 'Roberto Bordogna'; Roy
Roebuck; 'Walt Truszkowski'; 'Wu Hanxin '
Subject: RE: [COSMO-WG] UF starter kit: UF001 (06)
Adam,
In reference to: those who think that just aligning
ontologies after they've been created, rather than building on an
upper ontology from the start, is a sensible approach. (07)
Since every data structure, message and interface is, essentially, an
ontology, how would this be better than the current situation? You
would
have to align every one with every other one of interest - just like any
un-architected solution. And, there is KNOW WAY "normal" people making
schema and messages are going to define every term down to the precision
that would be required for any reliable automated adaptation. I Just
don't
see how it would work? (08)
Pat,
First - great work, Thanks! Getting something on the table is
invaluable.
Question - why do we need to start over, couldn't this group adopt one
of
the existing works as a baseline? (09)
Regards,
Cory Casanave (010)
-----Original Message-----
From: Adam Pease [mailto:apease@xxxxxxxxxxxxxxxxxxxxxx]
Sent: Wednesday, November 30, 2005 12:14 PM
To: Cassidy, Patrick J.; Anne Cregan; Antoinette Arsic; Arun Majumdar;
Barry
Smith; Brand Niemann; Charles Turnitsa; Christopher Spottiswoode; Cory
Casanave; Dagobert Soergel; David Eddy; Eric Peterson; Gary Berg-Cross;
Harold Frisch; I.N. Sarkar ; J. P. Morgenthal; James Schoening; Jeffrey
A.
Schiffel; John F. Sowa; John Thompson; John Young; Leo J. Obrst; Lowell
Vizenor ; Matthew R. West; Michael Denny; Michael Gruninger; Nicolas
Rouquette; Olivier Bodenreider; Patrick Cassidy; Pawel Garbacz; Peter
Yim;
Richard Lee ; Richard Murphy; Roberto Bordogna; Roy Roebuck; Walt
Truszkowski; Wu Hanxin
Subject: Re: [COSMO-WG] UF starter kit: UF001 (011)
Pat,
I wish people would be more cautious before asserting that some
concept or other doesn't exist in SUMO. Part of the problem is that
you haven't defined exactly what you think might be missing, other
than giving it an English phrase. But an "abstract informational
thing" could reasonably be assumed to be covered by the existing
notion of Proposition in SUMO.
A separate question is whether some particular notion is really
required. I recall discussing abstract character strings and similar
issues with Fritz Lehmann at some length years ago. From an academic
standpoint I'm very attracted to such a notion, but it hasn't made it
into SUMO so far simply because we couldn't identify a practical
need. Maybe some new project will change this.
Another issue is that because you've used OpenCyc, any mapping is
very problematic because it lacks rules that define the
concepts. You've had to guess at meanings based on the documentation
strings and basic type information. One could have greater
confidence in successful mappings between SUMO and DOLCE due to their
richer definitions. I applaud your energy, and that you're doing
real work rather than just having discussion, but I'm very skeptical
of this effort, I'm afraid. Once you get down to the real nitty
gritty of attempting alignment inclusive of the formal axioms the
difficulties of this task will become more apparent I think, and
serve as a good example for those who think that just aligning
ontologies after they've been created, rather than building on an
upper ontology from the start, is a sensible approach. (012)
Adam (013)
At 09:48 PM 11/29/2005, Cassidy, Patrick J. wrote:
>COSMO-WG:
>
>The discussion of the Unified Framework (UF) suggests that there is
>some level of agreement that it will be useful to try to find a
>top-level hierarchy that will not preclude use of any of the existing
>upper ontologies, which might be linked to the UF in the manner of a
>lattice of theories. I believe it will be helpful to prepare some
>specific examples of what might serve as such a UF.
>
>To provide one example for discussion, I have created a top-level
>hierarchy of about 100 classes, using mostly Opencyc and SUMO classes
>and adding in some classes from DOLCE, and a few each from BFO and ISO
>15926. The resulting hierarchy is on the Wiki page:
>
>http://colab.cim3.net/cgi-bin/wiki.pl?CosmoWG/TopLevel
>
>
>The bare hierarchy is listed there, and none of the metaclasses in any
>of the ontologies are included. and where classes appear to be closely
>similar in intent (the same intended instances), they are listed on the
>same line. About 50 of the SUMO classes appeared to align with OpenCyc
>classes. Because the OpenCyc (2002 version) is more detailed at this
>level, it was used as the base namespace, and unqualified class names
>are from OpenCyc. The top level of the SUMO and DOLCE that was used is
>also on that page, for reference, but those starting hierarchies are
>not a direct part of this proposed hierarchy (call it UF001 for the
>moment).
>
>If anyone else wishes to pursue this line of investigation, I will put
>the documentation of those classes from the original ontologies on that
>page also.
>
>Feel free to add comments or make additional suggestions on that page.
>The classes added from DOLCE, BFO and ISO 15926 were only a few, in
>part because (1) those ontologies are smaller; (2)I don't understand
>them as well as Cyc and SUMO; and (3) because they were sufficiently
>different that they might need significant additional effort, and
>perhaps some additional more abstract classes, to find the proper
>integration points. Anyone who thinks that there were useful concepts
>left out of this minimal top level, please feel free to add them where
>appropriate, or raise the issue for discussion. The base categories
>suggested by Roy Roebuck appear to be at a more detailed level than
>most of those in this minimal set, and would likely be included in the
>next stage of elaboration.
>
>One significant divergence is in the use of 4-D objects in ISO15926.
>Only the top level of the object classes was added in this first cut,
>because there will be no class subsumption between the other ontologies
>and any of the ISO 15926 object classes. The relations of these
>classes to the other ontologies will have to be via semantic relations
>or axioms. The 4D subclasses can be added if anyone wishes to have
>them visible.
>
>This hierarchy UF001 might be used as a starting point for three
>possible purposes:
>(1) as a straw man for the UF, to determine if the alignment suggested
>here is valid, or if there are actually logical differences in spite of
>the similar intended meanings, sufficient to require different
>representations, or to cut back the number of classes in the UF to a
>smaller number that might have more exact alignment of meaning.
>
>(2) As an aid in discovering the proper relations of the top levels of
>the major upper ontologies, which would be helpful in the work on
>formalization of the UMLS-SN or of the FEA-RMO or DoD core taxonomy.
>
>(3) As the beginning of a separate upper ontology (mostly a merger)
>with known relations to the existing upper ontologies and a wider range
>of ontological viewpoints, to be evaluated along with the others for
>utility in specific applications.
>
>======================
>This first quick alignment shows significant differences between Cyc
>and SUMO in a few points;
>
>(1) Abstract texts
> >>Generally, SUMO top lacks abstract representation of texts, a major
>difference from Cyc. As a result there are significant differences in
>these classes:
>
>Language
> in SUMO, a physical object containing linguistic symbols
> in Cyc, an AbstractInformationalThing, (but here are no abstract
>informational things in SUMO)
>
>Address
> Cyc - abstract string (no such thing in SUMO)
> SUMO treated as a relational attribute, with structure
>
>
>Possible fixes:
> In order to align more closely with the OpenCyc, a user could add
in
>abstract texts to SUMO (the Cyc hierarchy could be imported), or
>relations or functions can be added to create abstract texts from
>physical representations (or vice-versa).
>
>(2) Mental Objects - Descriptions
>
>In the DOLCE + Descriptions and Situations there is some more detail at
>a high level about mental concepts -- non-physical things that are the
>products of thought -- than in the other ontologies. These are
>important for representing motivation, goals, intentions, beliefs, etc.
>
>Possible fix;
> There appear to be correspondences in some places, but it seems
>likely that a significant expansion of detail in the mental area will
>be needed to properly represent all related concepts in each of these
>ontologies. Unless one or more ontologies are modified, this may be
>one area that will be sufficiently different to require separate
>microtheories.
>
>(3) Number and quantity
> There are slight differences between Cyc and SUMO, particularly in
>the relation of quantities to measures (e.g. 6 feet).
> At this point I cannot tell if these will be logically incompatible
>or may be reconciled with the proper relationships and axioms.
>
>(4) Partially tangible etc.
> The Cyc classes 'PartiallyTangible' and 'PartiallyIntangible' appear
>to be used to permit the use of a class of
>'CompositeTangibleAndIntangibleObject'. That last class in other
>ontologies is served by physical objects with certain 'intangible'
>properties. Relations that use these Cyc classes should be easily
>accommodated with axiomatic or functional transformations.
>
>---
>Otherwise both the intended meanings and the logical inheritance
>patterns appear close for most of the top classes, with some
>distinctions made in one that are not in the other, but not affecting
>the logical properties. Of course, the match isn't exact, and the
>logical differences will be more apparent when the relations are
>considered. But the UF001 may serve to permit more focused discussion
>of just how much difference exists among the existing upper ontologies.
>
>Other approaches may serve as well or better. Please feel encouraged
>to provide specific suggestions for such alternatives. It would
>probably be useful to create separate WIKI pages for each different
>approach.
>
>Pat
>
>Patrick Cassidy
>MITRE Corporation
>260 Industrial Way
>Eatontown, NJ 07724
>Mail Stop: MNJE
>Phone: 732-578-6340
>Cell: 908-565-4053
>Fax: 732-578-6012
>Email: pcassidy at mitre.org (014)
----------------------------
Adam Pease
http://www.ontologyportal.org - Free ontologies and tools (015)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (016)
|