Towards Natural Clarification Questions in Dialogue Systems

Abstract. Clarifications are often necessary for maintaining human-human as well as human-machine dialogue. However, clarifi-cation questions asked by Spoken Dialogue Systems (SDS) are very different from clarification questions asked in natural human interac-tion. While in human-human dialogues, speakers ask targeted ques-tions using contextual information, SDS ask generic clarifications such as please repeat or please rephrase. We propose and evaluate a new strategy for creating more natural clarification questions. We model natural language clarification question generation rules based on human-generated behavior. We describe results of a user study to evaluate our automatically generated questions and show that subjec-tive scores of the automatically generated questions are comparable to scores for human-generated questions — with some of the auto-matic rules even outperforming human-generated questions.