John Sowa has, I believe quite perceptively, pinpointed the methodology for
development of a common foundation ontology as one of the key factors that
could lead to success or failure. Perhaps it is worthwhile to give that
some more detailed consideration. (01)
To reiterate:
>
> PC> Well, the ideal foundation ontology (to serve as a means of
> > integrating independently developed domain ontologies) would have
> > in it any primitive concepts that are required to logically specify
> > the meanings of those domain concepts, and would be developed and
> > maintained by an open process involving the entire community of
> users.
>
[JS] > That's a fine goal. But we don't yet have anything that satisfies
> those requirements, and there is no structure in place for generating
> something that would.
>
>
> . . . But we don't have such a
> foundation that people can agree on. If some large organization
> pours millions of dollars or euros into a project to design one,
> there are several possible outcomes:
>
> 1. The project is hijacked by people who take the money and run.
>
> 2. Some pointy-haired boss takes charge of the project. (See the
> Dilbert cartoons for painful examples of what that leads to.)
>
> 3. Very talented people (such as us, for example) join the project,
> and we are assigned to a committee that produces as much results
> as we have seen from the SUO and ontolog email lists since the
> year 2000.
>
> 4. God sends an omniscient savior to earth, who personally chooses
> the best experts to design the best possible ontology that can
> solve all the world's problems.
>
> I won't hold my breath waiting for option #4.
> (02)
Nor would I. But I think we can design a process without divine
intervention that avoids the dire scenarios that John suggests as potential
outcomes. I will suggest one possible set of procedures that might be
adopted by a group developing the common foundation ontology. If a
significant group of potential participants can agree on a porcess, that
might be viewed as part of a proposal for funding. This assumes that the
funding requested would be adequate to support significant representation
from each interested community. (03)
If any others have suggestions for alternatives or modifications to this
process, we can work together to make it more plausibly effective. (04)
A funded project to develop a common FO would have an initial period of
about three years to develop the FO together with some aplications that
interoperate using the FO and an initial natural-language interface. Clear
goals and a clear timetable are essential for the initial project. A period
of several years after that will require some lesser funding for
maintenance. The FO will not likely become stable enough until several
years of use by a wider group of application developers. But I would expect
it to be stable enough to start thinking about making it a formal standard
in less than ten years. (05)
Keep in mind that byzantine rules may be necessary to both encourage and
take advantage of the widest possible participation while making progress as
rapidly as possible. Nevertheless designing a process that can accommodate
the interests of multiple groups should be a lot less complex than building
and testing the FO and its domain extensions. (06)
Basic principles, agreed to in advance by all funded participants:
1. The products of the project will be open and freely available to the
public at all stages, as soon as they become available for use by the
participants. Use is free, including commercial use. All contributions can
be considered as being in the public domain.
2. Participation is open to the public, but the paid participants will have
main control over the project.
3. Final decisions resolving any disagreements are taken by a majority vote
of the executive committee. Decision must be rapid, typically within a few
days; consensus is the goal, but when not possible, all participants will be
aware that their preferred views may not prevail.
3.1 The executive committee is elected by some combination of:
3.1.1 a majority vote of the paid participants
3.1.2 representatives of specific communities of interest (see below)
3.1.3 some representation of unpaid registered participants
3.2 The size of the executive committee will be decided by a vote of the
paid participants.
4. The voting process assures that no one or two participants wields a veto
over inclusion or exclusion of any component.
5. The goal is to include every element that any participant wants in the
ontology, either as part of the Foundation Ontology itself, or part of an
extension whose elements are logically specified using only the FO
components. The FO itself should be logically consistent, therefore only
one of any group of logically incompatible elements can be accommodated in
the FO. Others can be accommodated in extensions to the FO. Participants
should be aware from prelliminary discussions whether their favorite
representations are incompatible with others, and if so whether theirs may
not be allowed in the FO, but assigned to an extension. If the latter fate
is unacceptable, participants may be allowed to resign, foregoing further
funding and being released from obligations other than a final report on
their own activities.
6. Participants who are funded will agree to use the FO in some
application, and some applications must be designed to interoperate
semantically with others. Applications will be freely available for
download and local use, and some demonstration, particularly of the natural
language interface, must be available online for easy testing by the public.
7; Communities of Interest:
To function as a support for wide semantic interoperability, the FO
devleopment should include representatives from any potential user group,
including:
7.1 Relational Database Developers. To relate the logical models used to
generate an RDB, and even better to automatically generate an RDB from the
ontology, RDB developers will need to be included to be sure that the FO
will support that application.
7.2 Natural Language Understanding program developers. An effective NL
interface is needed to allow potential users to write definitions of domain
data elements or temrs in natural language (initially restricted English,
expanding over time), and have them automatically converted into the logical
representation. Similarly, users need to be able to query the ontology via
NL to determine whether a concept they want to use is already in the FO or
some extension.
7.3 Developers of reasoning systems, DL or FOL. There should be one
reference FOL reasoner to be used with the ontology, to illustrate how
specific logical structures (such as forall-exists rules) msut be handled,
so as to assure that the local reasoners all arrive at the same inferences
from the same data.
7.4 Developers of significant ontologies or other knowledge
representations systems such as the existing FOs, the NIEM, WordNet, the
Japanese EDR, the Chinese HowNet and Hantology, The OBO ontology developers,
public military ontologies such as the JC3IEDM, clinical systems such as
SNOMED or HL7 or the FMA, and several European ontology programs.
Developers of other Knowledge Organization systems such as the NCI
thesaurus, DoD Core taxonomy, the LOC classification system, and the Getty
Llibraray Thesaurus should be contacted to determine whether they would
anticioate benefit from being able to translate information to and from
other KOS's.
7.5 The W3C and the OMG.
8. Each identifiable community of interest may need at least one
representative on the executive committee, and this may affect the necessary
size of the committee. (07)
Suggest alternatives? (08)
Pat (09)
Patrick Cassidy
MICRA, Inc.
908-561-3416
cell: 908-565-4053
cassidy@xxxxxxxxx (010)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (011)
|