Re: A place for everything and everything in its place

Denys Duchier ([email protected])
Thu, 17 Mar 1994 21:40:15 --100


Hi Stan, long time no see.

[This is a reply to Stan Letovsky's message to www-talk.
I am CC'ing to some people who might be interested]

I am a bit pressed by time, so I'll have to keep it brief. Part of
the redundancy problem is already addressed by URNs. Now, I'll
restrict your proposed usage of the term "fact" to only denote a
statement associating a URN with a classification category, or one
associating one category with another. Possible applications are for
stating that a document is an element of a category, or that category
A is a sub-category of category B, or other forms of such
relationships. I have used the generic term categories to avoid
stronger connotations, but they could stand for topics, or taxonomic,
meronomic, or in fact arbitrary hierarchies. Which brings us to the
subject of semantic networks. I think that you are on the right
track, but that you should also look to the knowledge acquisition
community. Much work has been done on the representation and sharing
of ontologies (i.e. Genesereth etal, Gruber, Lenat^H^H^H^H^H :-),
Porter, Skuce, etc...). I am most familiar with the work of Doug
Skuce and Tim Lethbridge here at the University of Ottawa, and it is
no accident that I used the term "statement" earlier as this is the
basis of their approach. One of its nice features is that a statement
is a concept just like the property it is about and the subject and
object to which it applies. All concept items can have meta-concepts
to express knowledge about them: possible applications might be
expiration date, identity of maintainer, editorial judgments and
evaluation, etc. The problem you raise about the coordination of
"keyword assignment" would also be present, only more so, in a
distributed knowledge representation. The solution you propose,
namely that of designating curators for parts of the topic hierarchy
does not address the need to maintain concurrently a plurality of
views of the shared knowledge space. The interpedia project has
proposed the notion of SOAP - Seals Of APproval - that would allow
multiple "Editors" to express their endorsement of articles
contributed to the interpedia. A similar scheme can be extended to
apply to statements as described above. By subscribing to a subset of
these seals you would perceive the shared knowledge space as being
structured by a combination of their views. I cannot elaborate on
these ideas at this time...But I'll be back!

--Denys

PS: connecting knowledge bases to the net is not just a pipe dream; I
have put together a preliminary prototype WWW interface to a Code4
knowledge server which students have been using to explore existing
ontologies through Mosaic.