Dynamic ontology evolution in open environments

Abstract

Interoperation between knowledge-based systems or agents requires common ontologies to facilitate successful information exchange. However, the openness of the Semantic Web means that the notion of there being common domain ontologies sufficient to cater for the requirements of a diverse range of consumers and producers of services has become untenable. In these types of environments it is necessary to consider that no ontology can be expected to remain unchanged throughout its lifetime. However, the dynamism and the large scale of the environment prevent the use of traditional ontology evolution techniques, where changes are mediated by a knowledge engineer [3]. We argue that the ability to estimate the impact of change a priori, i.e. before performing the change itself, is crucial, since this estimate can be used to assess the usefulness of the change. We assume that agents are capable of rational behaviour, and that they decide whether to change the ontology they commit to if the cost of the change (in terms of reclassification of knowledge) is offset by the benefits derived from the ability of a system to acquire new capabilities and therefore to achieve new tasks (or answer new queries, in the case of knowledge based systems). However, the agent’s decision making process follows the principle of bounded rationality [5]: agents operate with limited computational resources, and with partial knowledge of the environment [4]. We present an approach that evaluates the impact of change on an ontology a priori, without using reasoning, by estimating which set of axioms in an ontology is impacted by the change.

More information

Identifiers

Catalogue record

Export record