The sentences uttered by naive users of a spoken dialogue system greatly differ according to the dialogue context in which they are uttered. In this study, we first analyze the factors that differentiate sentences (linguistic complexity, distribution of disfluencies, etc.). Second, we investigate the effect of using a specialized (dialogue-context dependent) language model during recognition. Results suggest that improvements depend on the type of language model employed and on the sentence complexity.
Cite as: Gerbino, E., Baggia, P., Giachin, E., Rullent, C. (1995) Analysis and evaluation of spontaneous speech utterances in focused dialogue contexts. Proc. ESCA Workshop on Spoken Dialogue Systems, 185-188
@inproceedings{gerbino95_sds,
author={Elisabetta Gerbino and Paolo Baggia and Egidio Giachin and Claudio Rullent},
title={{Analysis and evaluation of spontaneous speech utterances in focused dialogue contexts}},
year=1995,
booktitle={Proc. ESCA Workshop on Spoken Dialogue Systems},
pages={185--188}
}