Olga - a conversational agent with gestures Multimodal Agent Interfaces

The Olga project has developed an animated agent interface for information services. The interface combines a graphical interface, spoken dialogue and an animated 3D ‘human-like’ character for multimodal interaction with users. The interaction is intelligently managed using techniques derived from spoken dialogue but extended for the graphical modality. The Olga agent is innovative in combining an interactive spoken dialogue system with a 3-D animated character using lip-synchronized synthetic speech and gesturing. Particular attention has been paid to ensuring that the behaviour of the agent is immediately comprehensible for the user. Synchronizing speech with mouth movements increases intelligibility, while facial expressions and gesturing realize the agent’s internal states and focus of the dialogue. can express feedback either independently or in conjunction with speech. By representing the agent as an animated character we gain access to these modalities so as to give the system a more apparent personality, hopefully making the experience more like a human interaction. However, it is essential that there is a reasonable match between the agent’s capabilities and its realization: the more human-like the realization, the more users expect it to have human-like capabilities in, for example, speech and language understanding. Consequently, while our animated character is realized with lip-synchronized speech and gesturing, it is also cartoon-like so as to elicit behaviour appropriate to the current state of interface technology.