Finalized Human-Computer Dialogue loses its efficiency when the system cannot give the relevant and expected answer. Then the real dialogue begins. The aim of dialogue is to reach a successful conclusion. But the real problem is to manage the misunderstanding and not the understanding. The aim of this study is to provide the system with means to recover communication failures. So we collected all the linguistic forms of failures and the strategies to avoid communication failures in our corpus. We attempt to propose a formalization of them. The corpus we worked on was gathered for the ESPRIT 3 European project called "INTUITIVE" (INTeractive Users Interface and Tools for Information in a Visual Environment) whose objective is the design and implementation of a multimodal query system including speech as input regarding the maritime domain. The INTUITIVE maritime demonstrator is intended to be a system enabling the commanding crew of a ship faced with a critical situation to access relevant information. The corpus analysed below consists of 13 oral dialogues between ship-masters, who will be the future users of the system, and an experimenter who provides the information they need to perform a task (extinguish a fire).
Cite as: Taleb, L., Luzzati, D. (1995) Finalized spoken dialogue modelling based on communication failure. Proc. ESCA Workshop on Spoken Dialogue Systems, 37-40
@inproceedings{taleb95_sds,
author={Latifa Taleb and Daniel Luzzati},
title={{Finalized spoken dialogue modelling based on communication failure}},
year=1995,
booktitle={Proc. ESCA Workshop on Spoken Dialogue Systems},
pages={37--40}
}