Contextual affect modeling and detection in interactive text-based dramatic improvisation

Real-time contextual affect detection from open-ended text-based dialogue is challenging but essential for the building of effective intelligent user interfaces. In this paper, we focus on context-based affect detection using emotion modeling in personal and social communication contexts. Bayesian networks are used for the prediction of the improvisational mood of a particular character and supervised & unsupervised neural networks are employed respectively for the deduction of the emotional indications in the most related interaction contexts and emotional influence towards the current speaking character. Evaluation results of our contextual affect detection using the above approaches are provided. Generally our new developments outperform other previous attempts for contextual affect analysis. Our work contributes to the journal themes on emotion design/modeling for interactive storytelling, narrative in digital games and development of affect inspired believable virtual characters.