Efficient cultural models of verbal behavior for communicative agents

Abstract

This paper presents compositional models of culture for conversational agents that are embedded in a training system for cross-cultural competency. Our models are implemented as ontologies of statements in common logic, with culture-specific and culture-general components. We compare the compositional framework to a finite-state system, in terms of development effort, number of reused and new objects, and flexibility and accuracy of resulting conversational simulations. © 2012 Springer-Verlag Berlin Heidelberg.