Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: SUO: Program wants to work with SUO documents




Jim --
    Below is my suggestions for questions that can be
applied to upper ontologies for potential use.

    First, however, to comment on potential military uses:

(1)  Any ontology that is intended to be public will be available
     for use  by military organizations.  The IEEE-SUO is intended
     to be public.
(2)  I did not interpret Jim's request as soliciting any formal response
  by the SUO group as a group, but as a request to individuals with
  expertise in ontology to help define evaluation criteria for existing
  ontologies.  Those criteria may have immediate use for military
  organizations, but they will also be of use to any organization
  that is considering using an upper ontology in their
  applications.  The latter is entirely appropriate for a group
  concerned with standards development.
(3) I respect the views of those who are reluctant to see any of
  their work used by the military.  I am sure that they are
  aware that the military may use any public information, but
  if even an expression of interest by the military is sufficient
  to create a moral dilemma, they can chose not to participate
  in the creation of these evaluation criteria while continuing any
  positive contributions to a general standard.  It would be
  regrettable if explicit mention of potential military use
  would cause any members to cancel participation.  That
  would harm potential civilian uses perhaps even more than
  the military, as civilian applications are more susceptible
  to interoperability problems.

===================================================

My suggestions for evaluating potential upper ontologies:

(1) Are there any freely available applications that use the
   ontology and can be modified and tested locally to see
   if the ontology is suitable for one's own application?

   If not free, are such applications available for a modest
   fee ( < $100)?

(2) If not, are there proprietary applications for which
  examples of results in several problem areas are available
  that would serve to allow others to evaluate the ontology?

(3) If demonstration results are not available, are there
proprietary applications that use the ontology in any area?
Which area?

(4) Internal criteria of design and breadth.

   (4a)  How much time (for how many people) would it take
   for an outside group to adapt the ontology to a novel use:

      1. without direct assistance from the developing group?
      2  with funded assistance from the developers?

   On how many examples of experience is the estimate based?


   (4b) Potential for improvement: is the ontology available
  in a form that can be understood by domain experts with
  minimal (< 1 week) training so that they can add knowledge
  directly?  Or will it have to be extended by experienced
  ontologists?

    Is there a specialized development environment to
  ease the effort of knowledge entry?

    If the ontology is proprietary, will the users be able to
  modify it and use it without further cost, or will maintenance
  or expansion require additional fees?


  (4c) Breadth: How well will it support many domains?
       How many domain extensions (requiring more than 200
      specialized concepts) have been built using the
      upper ontology?  Which domains?  How many of these
      have been directly tested in applications?

  (4c) Language: In what representation language is the
   ontology being developed?  Is there an automatic converter
   into any other representation language?  Is the full
   development language specification freely available to the
   public?  Does the language support full first-order
   inference or is it a restricted subset of first-order
   logic?

     Are there specialized reasoning modules available?  How
   many and of what capability?

  (4d) Current size: how many concepts are included in the
    present ontology:
      classes?
      relations (slots/roles/attributes/associations)?
      axioms (logical implications)?
      instances?
      real-world assertions?


(5)  Alternate versions
   In the case of systems, like CYC or SUMO, that are
available in a smaller public form, the above questions
would have to be answered for one or both of the
systems individually.  The capabilities of a full
proprietary system cannot be attributed to a small
portion released publicly.

   ======================================================

   Comment:
    form my own understanding of CYC and SUMO as well as lesser
acquaintance with other ontologies, I would suggest that
any serious ontology development effort with a deadline of
more than two years not adopt any existing ontology, but
use a part of the necessary adaptive effort to create a
broader and more comprehensive upper ontology by including
all useful and consistent elements from all existing
ontologies.

    Pat Cassidy

-- 
=============================================
Patrick Cassidy

MICRA, Inc.                      || (908) 561-3416
735 Belvidere Ave.               || (908) 668-5252 (if no answer)
Plainfield, NJ 07062-2054        || (908) 668-5904 (fax)
				
internet:   cassidy@micra.com
=============================================