Message 11841 of the SUO list   (source: http://web.archive.org/web/20110806110851/http://suo.ieee.org/email/msg11841.html)

Subject: Re: Ontology Registry
Date: Wed, 26 Nov 2003 11:45:37 -0500 (Thu 02:45 EST)
From: "John F. Sowa" <sowa@bestweb.net>
To: cg@cs.uah.edu, SUO <standard-upper-ontology@ieee.org>
In reply to: my comments about John Sowa's answer to msg11829 by Bill Andersen
Follow-up: msg11936 by Philippe Martin forwarded by Jim Schoening


Philippe,

I believe you have made a major contribution to implementing
what everybody else has been talking about.  And I also think
that your approach helps to flesh out and extend the registry
proposal:

 > ... For me, all the categories of all ontologies have "some"
 > value (since each of those categories has been created and
 > represent something) and, when not identical, are complementary.

My observation that "Everybody else's upper ontology is totally
unacceptable" was meant in the sense that none of the ontology
developers would be willing to accept anybody else's upper
ontology as the foundation for their own categorial scheme.

However, I also believe that the *distinctions* that underlie
any or all of the schemes are more important than the individual
categories in them, and I have recommended that we build the
upper ontology on the basis of the distinctions.  If all of the
useful distinctions from each of the ontologies were selected as
the basis, then it should be possible to have a single hierarchy
in which each category from each of the starting ontologies would
find its natural position.

You have done the really hard work of making those selections and
implementing them in an actual hierarchy:

 > I merged various ontologies (John's, Dolce, WordNet, ...) into a
 > "multi-source ontology" (i.e. with each category identifier including
 > an identifier of the source/creator) for the shared ontology of my
 > knowledge server (www.webkb.org) and Web users can insert new
 > categories.  During the merge, I admit that I had to break a few
 > subsumption/exclusion links between WordNet categories but only
 > because those links led to inconsistencies (WordNet is inconsistent).
 > I have not broken links from the other ontologies I re-used.

I believe that this is a very important step toward a usable upper
ontology.  The next step would be for the developers of each of the
source ontologies to examine your merger and to satisfy themselves
that your hierarchy correctly represents their intentions -- and
if not, to suggest any corrections that may be necessary.

 > One of the problems of a registery of ontologies (as in the Ontolingua
 > server) compared to a multi-source ontology (as in WebKB) is that it
 > is difficult for an ontology provider to relate the new categories (by
 > subsumption/exclusion/identity/... links) to the categories of all
 > other ontologies in the registry, and hence these ontologies are
 > difficult to compare and re-use: each user must select various
 > ontologies (and choose between competing ones) then complement and
 > inter-relate their categories which is even more difficult than it
 > would have been for the authors of the selected ontologies.

I agree.  We certainly need something more than just a registry as
in Ontolingua.  What you have accomplished is what I was originally
proposing:  a selection of modules, each of which was independent, but
each of which was related to the others by their mapping to a super-
hierarchy of categories that included all the categories from each
of them.  My original proposal was very sketchy, but it outlined
something along the lines that you have actually implemented.
I believe that your multi-source ontology is something that we
should very seriously consider as a foundation for the SUO.

 > Conversely, with a multi-source ontology server, (i) it is easy to
 > access categories and see/navigate their inter-relations, and hence
 > insert/relate a new category to the other ones (this process also
 > limits redundancy and leads to the creation of more precise and
 > re-usable categories), (ii) if needed, it is easy to filter out all
 > the categories that are not used by an application (i.e. the required
 > minimal ontologies can be generated, they do not have to be found,
 > selected, merged and complemented; the inclusion relations between
 > these ontologies may also be generated although I do not see their
 > interest).

 > Although a multi-source ontology and a registery/lattice of ontologies
 > are semantically equivalent, I fail to see how the latter can, in
 > practice, support "scalable" knowledge sharing.

I believe that your superhierarchy of categories is a necessary part
of any suitable, scalable knowledge sharing effort.  And your filters
make it possible to have multiple views of the ontology.  For example,
you could push one button to select all categories from Dolce, and
you would see the original Dolce ontology by itself.  Then you could
push a different button to see all categories from both Dolce and
SUMO -- the result would be a merged ontology of SUMO + Dolce.  By
pushing more buttons, you could see any merger of any combination
of the source ontologies.

Since you have that superhierarchy, it is possible for the filters
to pull in or screen out any contribution from any of the sources
that went into its construction.  But the registry ideas should
also be included:  each module by itself should be documented and
annotated with all the information about its history of development,
contributors, testers, and epecially all significant applications.

>  > Would it not be more fruitful to ask those who have "been there
>>  > and done that" whether top levels are useful and which features
>>  > they have found useful through experience?
>> 
>> You can ask, but not many people have given clear answers.
> 
> Some interests of top-level categories are
> (i) they structure the ontology,
> (ii) they are needed for the signature of 
>      spatial/temporal/thematic/rhetorical/... relations,
> (iii) they have associated partial/total definitions and schemas or
>       prototypes, which are inherited by their specializations, and
>       hence permit to check or guide knowledge entering.

I recommend that your multi-source ontology be adopted as a starter
document for the SUO ontology.  I believe that it is the most
significant contribution we have seen so far that demonstrates how
multiple source ontologies can be shared, coordinated, and reused.

John Sowa