Affiliations: Department of Computer and Information Science,
Linköping University, S-581 83, Linköping, Sweden | The College of Economics and Computer Science,
Olsztyn, Poland
Abstract: Currently, there is a great deal of interest in developing tools for
the generation and use of ontologies on the WWW. These knowledge structures are
considered essential to the success of the semantic web, the next phase in the
evolution of the WWW. Much recent work with ontologies assumes that the
concepts used as building blocks are crisp as opposed to approximate. It is a
premise of this paper that approximate concepts and ontologies will become
increasingly more important as the semantic web becomes a reality. We propose a
framework for specifying, generating and using approximate ontologies. More
specifically, (1) a formal framework for defining approximate concepts,
ontologies and operations on approximate concepts and ontologies is presented.
The framework is based on intuitions from rough set theory; (2) algorithms for
automatically generating approximate ontologies from traditional crisp
ontologies or from large data sets together with additional knowledge are
presented. The knowledge will generally be related to similarity measurements
between individual objects in the data sets, or constraints of a logical nature
which rule out particular constellations of concepts and dependencies in
generated ontologies. The techniques for generating approximate ontologies are
parameterizable. The paper provides specific instantiations and examples.
