Finding a Meta-dialogue Strategy for Multi-intent Spoken Dialogue Systems

摘要

Speech is an easily accessible and highly intuitive modality of communication for humans. Maybe that is the reason why people have wanted to talk to computers almost from the moment the first computer was invented. Today several consumer-level products developed in the last few years have brought inexpensive voice assistants into everyday use. The problem is that this speech interfaces are mostly designed for certain commands. But during demanding tasks like driving a car, it can be useful to talk about several things at once, to get back to the main task as fast as possible. While talking about different things in a single utterance it is important to give the user adequate feedback, like a meta-dialogue informing about which topic is discussed at the moment. In this paper we compare several meta-dialogue approaches for a speech dialogue system capable of handling multi-intents. The aim of our study is to reveal which strategies users prefer regarding metrics such as flexibility, joy, and consistency. Our results show that explaining topic transitions and topic introductions via speech receive a high user rating and is cognitively less demanding then visual cues.