Copy the page URI to the clipboard
Dietze, Stefan and Tanasescu, Vlad
(2009).
URL: http://www.aisb.org.uk/convention/aisb09/Proceedin...
Abstract
The increasing availability of ontologies raises the need to establish relationships and make inferences across heterogeneous knowledge models. The approach proposed and supported by knowledge representation standards consists in establishing formal symbolic descriptions of a conceptualisation, which, it has been argued, lack grounding and are not expressive enough to allow to identify relations across separate ontologies. Ontology mapping approaches address this issue by exploiting structural or linguistic similarities between symbolic entities, which is costly, error-prone, and in most cases lack cognitive soundness. We argue that knowledge representation paradigms should have a better support for similarity and propose two distinct approaches to achieve it. We first present a representational approach which allows to ground symbolic ontologies by using Conceptual Spaces (CS), allowing for automated computation of similarities between instances across ontologies. An alternative approach is presented, which considers symbolic entities as contextual interpretations of processes in spacetime or Differences. By becoming a process of interpretation, symbols acquire the same status as other processes in the world and can be described (tagged) as well, which allows the bottom-up production of meaning.