Clarification in Spoken Dialogue Systems

Natural Language Generation in Spoken and Written Dialogue

Volume

Papers from the 2003 AAAI Spring Symposium

Contents

This paper argues for the use of more natural and flexible clarification questions in spoken dialogue systems. Several forms and aspects of clarification questions are discussed and exemplified by examples from human-human dialogue corpora. We show how clarification questions fit into the broader context of miscommunication and describe the types of information spoken dialogue systems need in order to generate various different clarification forms. Two kinds of dialogue systems are considered: Slot-based dialogue systems and dialogue systems with deep semantic analysis.