Error Handling in Spoken Dialogue Systems August 28-31, 2003 Château d'Oex, Vaud, Switzerland

User States, User Strategies, and System Performance: How to Match the One with the Other

Anton Batliner (1), Christian Hacker (1), Stefan Steidl (1), Elmar Nöth (1), Jürgen Haas (2)

(1) Chair for Pattern Recognition, University of Erlangen-Nuremberg, Germany (2) Sympalog Voice Solutions GmbH, Erlangen, Germany

Apart from the ‘normal’ linguistic information entailed in user utterances - segmental (phone/word) information and syntactic/semantic information - there is additional information (supra-segmental and para-linguistic) that can be useful for deciding whether an automatic dialogue system performs well or not. In this paper, we want to deal with such additional information and correlate it with system performance. Moreover, we will examine whether prosodic peculiarities influence word recognition.

Full Paper

Bibliographic reference.  Batliner, Anton / Hacker, Christian / Steidl, Stefan / Nöth, Elmar / Haas, Jürgen (2003): "User states, user strategies, and system performance: how to match the one with the other", In EHSD-2003, 5-10.