Dialogue Processing and System Involvement in Multimodal Task Dialogues

Abstract

We compare dialogue processing needs for several types of applications involving human-human and human-system collaboration and communication in various scenarios for a range of tasks. The tasks in these scenarios include collaboration which requires communication for information exchange and for task management. However, the necessary dialogue processing in the corresponding applications differs depending on the level of system involvement in the task(s) at hand, and in this paper we propose to put them on a continuum corresponding to the depth/coverage of the dialogue processing that is required. We provide a different and complementary perspective on ways of comparing dialogue processing than is common when looking at dialogue as either goal/task-oriented or chit-chat, or when considering different approaches to dialogue management, such as finite-state trans-ducers or template-filling. We explore the possibility of gradually scaling dialogue processing complexity while sharing development resources for different applications along the continuum within one application domain.