CONNECTIONIST REPRESENTATION AND DISCOURSE PROCESSING

Abstract

Trying to simulate human discourse processing, connectionism has developed several approaches in order to represent frames and scripts. These models are able to go far beyond of some of the shortcomings of classical systems. Specifically, classical systems are unable to control when a chunk of knowledge has to be activated, and with the issue of what part of the chunk should be more activated. Instead, connectionist systems admit to use activation degrees for dealing with this issue. However, connectionist models have also some limitations if considered as models of human mind. Finally, the plausibility of hybrid approaches is analysed.