Published: 2005 Juni
Type: SEKT Deliverable
Institution: University of Karlsruhe
The world is constantly changing, and so does required and available knowledge, e.g. stored in Digital Libraries. Knowledge workers heavily rely on the availability and accessibility of knowledge contained in such libraries. The sheer mass of knowledge available today, however, requires sophisticated support for searching and, often considered as equally important, personalization.
In SEKT we address these challenges by using ontologies. Ontologies by nature make implicit knowledge explicit, they describe relevant parts of the world and make them machine understandable and processable. To be effective, ontologies need to change possibly as fast as the parts of the world they describe.
Change discovery aims at generating implicit requirements by inducing ontology changes from existing data. In this deliverable we focus on usage-driven change discovery. Usage data is a very valuable source of contextual information, based on which the ontology can be modified in order to reflect changes in the real world,
We describe a framework, in which ontology evolution and
discovering potentially useful changes can be formalized as an
optimization problem. We introduce the notion of an evaluation
function that allows to measure the quality of an ontology with
respect to given criteria. We instantiate the framework for the
task of ontology pruning based on usage data, and for the task of
collaborative evolution in a multi-user scenario, in which users
maintain personalized ontologies. A first case study based on the
Bibster system shows very promising results.
Download: Media:2005_984_Haase_Usage Tracking _1.pdf
Weitere Informationen unter: Link