Evaluation of a mixed-initiative dialogue multimodal interface

This paper presents the speech recognition accuracy testing and usability evaluation of a mixed-initiative dialogue multimodal interface for the ATTAIN* travel information system [1]. Experimental results show that although the speech recognition accuracy of the interface is less (sample accuracy rate 74.5%) than that of an equivalent directed-dialogue interface (sample accuracy rate 88.5%), the usability is seen to be significantly improved in terms of effectiveness, efficiency, learnability and user satisfaction.