8.3313 Conceptual Spaces - Applications and Learning (KOGW-MWPM-CL) (KOGW-MWPM-KI) (KOGW-MWPM-IDK)Bechberger, Gregoromichelaki, Potyka
MSc: Major subject
MSc major: Linguistics and Computational Linguistics
MSc major: Artificial Intelligence
MSc Interdisciplinary Course
Conceptual Spaces (Gärdenfors 2000, 2014) is a framework aimed at bridging the gap between current symbolic (logic-based) theories of knowledge representation and non-symbolic ones, as in, e.g., artificial neural networks. From a cognitive science point of view, the interest lies in the potential to study the interaction between the data extracted from low-level stimuli involved in perception and the more abstract representations employed by higher cognitive processes. In this connection, Conceptual Spaces provides a basis for deriving formalisms combining these through mathematically well-founded, geometrically expressible information structures where abstract conceptualisations are modelled as multi-dimensional spaces whose dimensions correspond to interpretable qualities (features) like colour or size and concepts correspond to geometrically coherent regions within these spaces. Cognitive scientists and linguists have seen such formalisms as a suitable way to capture notions like similarity, analogy, degree judgements, prototypical membership and the effects of context on these. From an AI perspective, we are interested in representing derived conceptual spaces in a computer, learning them from data, and performing reasoning operations on them. From a psycholinguistic perspective, conceptual spaces provide a way to model the online derivation of the contextually-appropriate ad hoc concepts that underpin our use of words, creative behaviours and reasoning, and the processes of adjusting individual conceptualisations in order to coordinate joint action.