Adaptive dialogue management for a script knowledge based conversational assistant
Isabel Feustel · OPen Access Repositorium der Universität Ulm (OPARU) (Ulm University) · 2019
The spoken dialogue systems today already fulfill many requirements, and their human-machine interaction works very well. Nevertheless, the users still adapt their communication style to these system. In order to develop appropriate communication style strategies for a spoken dialogue system, we have investigated the communication style factors elaborateness and directness. Within this work, an indoor navigation spoken dialogue system was developed, which has different communication style strategies and states of knowledge implemented. Moreover, a formula for the adaptation of the system's communication style to the user's communication style was developed. The system navigates the user through the University of Ulm based on scripts. In order to create these scripts, a data collection study was carried out in which 97 participants described the selected routes with the help of videos, which have been recorded beforehand. Finally, the system was evaluated in a user study with 30 participants. In summary, the system performed good, but we could find no significant difference between the investigated communication styles. However, the different knowledge states of the system revealed some significant differences. Overall, we were able to demonstrate the potential of the indoor navigation application and gained new insights for the communication styles and the interaction with a non-omniscient system.