Speech-based dialogue design for usability assessment

Rudnicky and Hauptmann (1992) have suggested that speech by itself may not be an ideal input modality, and they implicitly suggest that information should be redundantly coded in other modalities. It was, therefore, advantageous to develop a system which could explore different direct voice input/output (DVIO) dialogues and this was the reason for developing Dialogue Designer, a dialogue prototyping system. A major problem with currently available speech recognition systems is the small percentage of failed recognitions. Even though there are fewer and fewer failures, they have a significant impact on behaviour in time-critical applications. Curiously, it is in the realm of time-critical applications that DVIO has received the greatest interest, because of the potential reductions in workload afforded by transferring information delivery to an alternative modality. The authors feel that this strategy may be flawed in that operators using such systems may rely on their memory for auditory information and consequently make many more cognitive errors. Thus, a design tool is required specifically to examine the effectiveness of DVIO dialogues on overall task performance in multitasking. Dialogue Designer was originally designed for simple unimodal operation, but it accommodates quite complex dialogue structures like those of Rudnicky and Hauptmann, and it may be used with other media