Key questions on Common Upper Ontology
4/20/2004 6:54
AM
Note from Leo Obrst,
There are low-hanging "semantic" fruit that can
be addressed, and that's the premise of the Semantic Web: islands of semantics
that can be linked. Since we have hardly any semantics at all now, that is an extremely
laudable goal, and the RDF/OWL communities are trying to achieve that. Can full
semantic interoperability be achieved without an upper ontology (and here I
would say, without upper ontologies, since the emerging common view is of a
lattice of theories, i.e., modules, not a monolithic ontology): probably not. Another
premise of the Semantic Web is that better ontologies will "out" over
time, and folks will incorporate what is better representation or simply use
it. That is no doubt true to an extent. But I would extend that to include
"domain reference" and upper ontologies: the better ones will
"out" and be used.
It's also not the case that upper ontologies have been
under development for 20 years without success. Until the past 2-3 years, there
has been only the Cyc upper ontology and some specialized natural language
"upper ontologies" addressing NLP and machine translation
applications. An upper ontology is not an ontology of everything, and it does
not have to be monolithic itself, in fact, should NOT be. The WonderWeb
approach of ontology libraries is an example. In fact, it is only in the past
10 or so years that ontology engineering has emerged as a technical discipline,
crucially incorporating analyses from formal ontology in philosophy and formal
semantics in linguistics. That IS new, and hence the newer upper ontologies
incorporate these analyses/insights and their logical rigor, as least to some
degree.
Leo