In March 2003, the OMG issued an Ontology Definition Metamodel RFP asking for propositions of extensions to
UML 2.0
in order to permit UML users to create/import/export ontologies (especially
OWL ontologies) and hence
represent and share various kinds of knowledge.
[Note added in February 2005. The four submissions (summary here)
that have been made to this RFP are being merged and have led to other documents,
e.g. chronologically 1,
2 and
3.
As one can expect, these documents have the content/approach of classic UML-related
documentation which, in my opinion, is rather inadequate (irrelevant and uselessly complex)
for describing or guiding "knowledge" modelling (as opposed to data modelling).]
As a first step to answer this RFP, Kerry Raymond (ex DSTC, now QUT) and I have created KRM, a MOF2 metamodel derived from the WebKB-2 data model which itself is an extension and normalization of the Conceptual Graph model. The WebKB-2 model and notations are of a higher-level than other models and knowledge representation notations (e.g. FOPC-oriented notations like KIF, frame-based notations like Frame-Logics, and database-oriented notations like RDF/XML) and hence ease knowledge representation, sharing and exploitation (inferencing).
A second step is to explain how UML and its notations can be extended to cover a
large range of knowledge representation cases, and how it relates to other
languages (models plus notations). I have done this by
comparing these languages according to a range
of knowledge representation features.
Click here for an ontology example extracted from the
OWL guide and corrected.
In addition to language ontologies and notations, we also propose a general content ontology composed of a merge of various top-level ontologies and an extension and correction of the WordNet natural language ontology (see this article for details). Some schemas have been associated to some of the top-level categories (many more will be added). WordNet categories have been re-used and specialized in various examples/applications.
For scalability and knowledge use and re-use purposes, some ways to represent knowledge are better than others. For now, click on on the following links: lexical recommendations, logical/semantic and ontological recommendations.
Finally, knowledge sharing involves protocols for editing a shared KB, or techniques to merge independently developped ontologies (and then KBs). WebKB-2 has some shared KB edition protocols.
History of KRM
Examples 1 Examples 2