15th European Conference on Artificial Intelligence
  July 21-26 2002     Lyon     France  
   

ECAI-2002 Conference Paper

[PDF] [full paper] [prev] [tofc] [next]

Anything to Clarify? Report Your Parsing Ambiguities!

Kerstin Bücher, Michael Knorr, Bernd Ludwig

Experience with spoken dialogue systems and an in-depth analysis of human-computer dialogues has shown that one of the main reasons for failed interactions is the difficulty to integrate new utterances in the dialogue context. This paper addresses the issue of grounding utterances in task-oriented human-computer dialogues. It focuses on the aspect of handling ambiguities that result from the parsing of input from a speech recogniser. An approach to parsing spontaneous speech is presented that is capable of computing the origins of ambiguities. The parser first segments input into chunks and then analyses their dependency relations on the basis of case frames contained in a semantic lexicon. In this way, pragmatic constraints can be applied early while semantic representations are constructed. Ambiguities usually are handed over to the dialogue manager as n-best parses that are not related to each other. In our approach, ambiguity reports mark the differences between the semantics of the different readings. Disambiguation is either achieved by exploiting the application situation or by initiating clarification dialogues that are suitable for the dialogue situation. This approach is an important step towards flexible handling misunderstandings caused by speech recognition and parsing in human-computer dialogues.

Keywords: Discourse Modelling, Natural Language Processing

Citation: Kerstin Bücher, Michael Knorr, Bernd Ludwig: Anything to Clarify? Report Your Parsing Ambiguities!. In F. van Harmelen (ed.): ECAI2002, Proceedings of the 15th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2002, pp.465-469.


[prev] [tofc] [next]


ECAI-2002 is organised by the European Coordinating Committee for Artificial Intelligence (ECCAI) and hosted by the Université Claude Bernard and INSA, Lyon, on behalf of Association Française pour l'Intelligence Artificielle.