1. There was early consensus, and indeed a formal minute, that "the
standard" to be developed was not to be an agreed set of terms or concepts
but norms of construction and annotation to which any "upper level
ontology" should conform. At the same time , the group supported the
proposal of putting forth in the public domain a "reference ontology" that
would illustrate one possible embodiment of the standard.
2. There early emerged differences in point of view stemming from three
distinct but not mutually exclusive uses to which an upper level ontology
might be put: a) translation (Pangloss, EDR), b) logical inference (Cyc,
Mikrocosmos), c) information retrieval and analysis (WordNet, with a
cross-language dimension distinct from machine translation added by the
European experience). Each of these purposive orientations put its own
stamp on the kind of ontology sought. To accommodate the diversity of
applications, the notion of "merging" ontologies was more precisely termed
"aligning" them, as in John's note, in a way that might preserve the
distinctive competence of each. In an alignment of upper level terms, it
is never assumed that the nominal occurrence of the same terms in two or
three ontologies signifies agreement on import.
I should also mention that the discussion to date, as I understand it, has
not required taking as canonical a hard-and-fast line between logic, as an
axiomatized but otherwise contentless formalism, and ontology, as
supplying the missing content. Indeed, some discussants use "ontology" to
refer to an axiomatized system, while others speak of "logical content" as
having a meaning that goes beyond a truth-functional calculus.
Lee
On Sun, 3 May 1998, John F. Sowa wrote:
> Perhaps a bit of history might help to clarify or refocus this discusion:
>
> 1. During 1996 and 1997, there have been four two-day meetings on ontology,
> which have been held as "ad hoc" working sessions attached to the
> NCITS T2 meeting (which is working towards ANSI and ISO standards
> for conceptual schema modeling facilities (CSMF)). The ANSI people
> have recognized the importance of logic as a precise formalism for
> stating axioms and definitions and ontology as the study of the content
> to be represented in logic. They are primarily oriented towards
> practical, commercial applications, but there is a core of people
> (which includes me and the T2 chairman Tony Sarris) who appreciate
> the practical value of theory.
>
> 2. The ad hoc ontology meetings, which have been attended by about half
> of the people on the above cc list plus another 20 or so others,
> have included some very interesting discussions which have led to
> some serious work in trying to reconcile previously independent
> projects. Ed Hovy and Fritz Lehmann, in particular, have done
> an important piece of work in aligning many of the nodes of Cyc,
> WordNet and Pangloss. Ed and I did some collaboration in merging
> our lists of thematic roles (or case relations) to form a version
> that we have been using for labeling the participants of verbs.
> This work has been promising, but much more has to be done, and it
> should be done with public scrutiny by the best available talent in
> linguistics, logic, philosophy, lexicography, AI, and any discipline
> that has concepts to define that might be related to this work (which
> includes essentially everything).
>
> 3. In January of this year, we held a two-day planning meeting at CSLI
> for a meeting of experts in these fields to be hosted by the
> Klaus Tschira Foundation in Heidelberg. The major accomplishment
> at that meeting was to assemble a list of about 40 people we would like
> to invite, which we narrowed down to a primary list and a list of
> standbys. The cc list above is a result of this process.
>
> 4. We also discussed how we might proceed in such a meeting, and I don't
> believe that discussion was as thorough or as definitive as the list
> of people. The most definitive part was to plan several presentations
> or position papers for the first day. Another suggestion was to have
> a list of concepts that could serve as a test bed for analysis, but
> the exact question of how that testbed could, would, or should be
> used was not clarified by the end of the two-day meeting. Piek Vossen
> volunteered to contribute about 40 concepts that were common to all
> the European languages that went into EuroWordNet. I will leave it
> to Piek to explain that further.
>
> The people who have been attending the four two-day ontology meetings over
> the past two years have felt that they were making progress towards aligning
> independently developed ontologies. But there are still some very important
> methodological issues that remain to be analyzed further and be sujected to
> theoretical scrutiny. Equally important is the detailed work on the large
> ontologies that have been asssembled so far, such as Cyc, WordNet, and EDR.
> Representatives from all these groups have attended one or more of the
> meetings, and most of them believe that further progress is both possible
> and highly desirable.
>
> This is a brief summary of my impressions of what has happened. Perhaps
> some of the others who have attended the earlier meetings can add a few
> words about their views of what has happened and how we might proceed.
>
> John Sowa
>
>