Dramatic Expression in Opera, and Its Implications for Conversational Agents

摘要

It is commonly agreed among embodied conversational agent (ECA) researchers that ECA behavior should be based upon principles of human face-to- face communication (Cassell et al., 2000; Traum & Rickel, 2002). It is less commonly acknowledged that principles of human acting can inform the design of ECA behavior, particularly in making behavior engaging and understandable. Character animators, in contrast, understand clearly the relationship between character behavior and acting (Porter, 1997), and have articulated principles such as exaggeration and staging that are based in part on observations of actors (Thomas & Johnston, 1981; Lasseter, 1987; Maestri, 1999). However, we cannot expect to capture principles of dramatic portrayal in ECAs simply by copying the techniques of animators. ECAs are being developed for a applications with a variety of media characteristics; we therefore need to draw lessons from a range of dramatic media, including those involving live action. Some ECA developers try to incorporate dramatic aspects by collecting motion capture data from actors (Churchill et al., 2000). This approach relies upon the actor's expressive skills to achieve the desired dramatic effect. Unfortunately there is no assurance that motion capture data will appear equally expressive and appropriate when transferred to different media and different dramatic contexts.