Publikation

Techniques for Organizational Memory Information Systems

Andreas Abecker, Ansgar Bernardi, Knut Hinkelmann, Otto Kühn, Michael Sintek

DFKI DFKI Documents (D) 98-02 1998.

Abstrakt

The KnowMore project aims at providing active support to humans working on knowledge-intensive tasks. To this end the knowledge available in the modeled business processes or their incarnations in specific workflows shall be used to improve information handling. We present a representation formalism for knowledge-intensive tasks and the specification of its object-oriented realization. An operational semantics is sketched by specifying the basic functionality of the Knowledge Agent which works on the knowledge intensive task representation. The Knowledge Agent uses a meta-level description of all information sources available in the Organizational Memory. We discuss the main dimensions that such a description scheme must be designed along, namely information content, structure, and context. On top of relational database management systems, we basically realize deductive object-oriented modeling with a comfortable annotation facility. The concrete knowledge descriptions are obtained by configuring the generic formalism with ontologies which describe the required modeling dimensions. To support the access to documents, data, and formal knowledge in an Organizational Memory an integrated domain ontology and thesaurus is proposed which can be constructed semi-automatically by combining document-analysis and knowledge engineering methods. Thereby the costs for up-front knowledge engineering and the need to consult domain experts can be considerably reduced. We present an automatic thesaurus generation tool and show how it can be applied to build and enhance an integrated ontology/thesaurus. A first evaluation shows that the proposed method does indeed facilitate knowledge acquisition and maintenance of an organizational memory.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence