Anything to Clarify? Report Your Parsing Ambiguities!

An important factor for the acceptance of spoken dialogue systems is their ability to react flexibly on misunderstandings between user and system. This paper addresses the issue of grounding utterances in task-oriented human-computer dialogues. It focuses on the aspect of handling ambiguities while parsing word lattices from a speech recognizer: The parser detects the origin and the type of ambiguities which are reported to the dialogue manager as comments to the list of readings for the user’s utterance. This way, disambiguation is delegated to the dialogue manager and accomplished either by exploiting the application situation or by initiating clarification dialogues that are suitable for the dialogue situation.