ESSV Konferenz Elektronische Sprachsignalverarbeitung

Title: A toolkit for nested multi-turn speech dialog in automotive environments

Authors: Timo Sowa, Soyuj Kumar Sahoo

Abstract:

Current in-car dialog systems foster concise interactions for navigation,media, or phone control very well. Yet, they are usually not employed for complexmulti-turn dialogs involving nested topics. However, this type of conversation isgetting more important when moving toward more natural, assistant-like voice interfaces in the car. We present a dialog modeling and execution approach to accountfor this. It is based on the idea of a dialog stack that keeps track of the active dialogtopics. The stack influences the execution of dialog turns while executed dialogturns modify the stack. The approach supports dialog decomposition into reusableparts and topic interruption/resumption for user-initiative and mixed-initiative dialogs. We implemented our approach for an existing GUI and dialog modelingtool.


Year: 2019
In session: Poster und Demonstrationen
Pages: 231 to 238