This paper describes ongoing work which aims to produce a methodology for automatically deriving several prosody models, each suited for text-to-speech synthesis of a specific text type. As part of this work an experiment is described in which human readings of documents covering 10 common text types (news, weather forecasts, etc) were analysed for their acoustic/prosodic properties. This was carried out in three languages - Dutch (Belgium), English (USA) and French (France). The results confirm that there is substantial variation in prosody across different text types.
Cite as: Fackrell, J., Vereecken, H., Buhmann, J., Martens, J.-P., Coile, B.V. (2000) Prosodic variation with text type. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 3, 231-234, doi: 10.21437/ICSLP.2000-520
@inproceedings{fackrell00_icslp,
author={Justin Fackrell and Halewijn Vereecken and Jeska Buhmann and Jean-Pierre Martens and Bert Van Coile},
title={{Prosodic variation with text type}},
year=2000,
booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
pages={vol. 3, 231-234},
doi={10.21437/ICSLP.2000-520}
}