Affective Space Calibration in Action-rich Media Affective Systems

Affective computing systems are software systems that take in account the emotional state of the user during their operation. Recently it is increasingly accepted that combining emotional with intellectual processing makes a system both more realistic and more intelligent. As such, affective computing systems are considered by many to represent the highest form of artificial intelligence and the state of the art in computer human interaction. A major difficulty in building an affective computing system is to identify what emotion attributes are most applicable for the development of the particular system that is under development. If such information is available then a system’s developer can build the sought system by using only the emotion attributes that are most influential for that type of system. In this work we present a method that addresses this problem. We describe our method and present an evaluation from its implementation and experiments with two different media collections. Our findings indicate that the method can successfully pinpoint to the most influential emotion attributes and thus allow the prospective developer to select the most applicable attributes for building a production-grade affective computing system.