Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
People often talk about more than one topic at the same time. Sometimes, even several intents concerning different topics are realized in one utterance. Combining intents allows for effective and swift dialogue. These advantages in communication among humans can also be used in the interaction with a speech dialogue system, if it has the ability to understand utterances with multiple intents and to react adequately to them. Utterances with more than one intent that involve different activities underlying the dialogue are called multi-intents. The focus of current multi-intent research is on recognizing, segmenting, and understanding multi-intents. How to deal with these utterances in a speech dialogue with a machine has rarely been addressed. In this work, a dialogue strategy for dealing with multi-intents was developed. This strategy involves prioritization and sequentialization of the intents. Prioritiza- tion is based on urgency, which is determined using four factors such as acoustic features. Meta-dialogue elements are used to transition to the next intent and to signal to the user that more than one intent has been understood. In a user stu- dy could be shown that the developed dialogue strategy concept of an urgency- adaptive dialogue behaviour in contrast to a comparison system can be more effi- cient and is perceived better in terms of usability. The basis for the concept development was formed by the results of several in- vestigations with participants. For example, in a Wizard-of-Oz study it was shown that meta-dialog elements introducing the start of a new topic and thus prepare to- pic changes can reduce confusion and increase perceived consistency. In addition, a gamification experiment demonstrated the feasibility of urgency classification using acoustic features. From the resulting data set, a reliable classifier model for automatic urgency detection was developed. The feasibility of the whole concept was proven by implementing an urgency- adaptive speech dialogue system prototype in the context ...