Combining a Data-Driven and a Theory-Based Approach to Generate Culture-Dependent Behaviours for Virtual Characters

摘要: To incorporate culture into intelligent systems, there are two approaches that commonly proposed. Theory-based build computational models based on cultural theories to predict culture-dependent behaviours, and data-driven rely multimodal recordings of existing cultures. Based our former work, we present a hybrid approach integrating Bayesian Network aims at predicting non-verbal behaviours for given conversation. While the model is structured theoretical knowledge their influence prototypical behaviour, parameters learned from corpus recorded in German Japanese The validated ways: With cross-fold validation estimate power network by parts data were not used train network. Secondly performed perception study with virtual characters whose behaviour driven calculations rated members this chapter, aim giving guidance other culture-specific generation providing methodology as well potential evaluation.