An Investigation of Language Input and Performance Timing for Task Animation

We describe a prototype system in which task animation is driven via natural language. The primary effort in developing the system is concentrated on the link between the natural language parser and the animation environment. Two primary problems are object referencing and specifying action durations. We describe a technique by which objects referenced by the parser can be correctly mapped to their geometric representation within the animation environment even though the internal representations may be vastly different. Furthermore, we show that results from experiments measuring human motor behavior can be applied to computer simulations to generate default task durations.