Our study is concerned with the modelling of intonation in Swedish, primarily prose reading. It is basically a superposition model with accentuation patterns added to prosodic base contours with prescribed onset, offset and declination. Boundary conditions at a juncture are related to sentence type and pause duration. There are several novel features of our approach. It is based on analysis of F0 data on a semitone scale of each syllable of a text read by five subjects of which two females. A sampling convention inspired by the canonical accent 1 and accent 2 typology of Bruce [1] is adopted. Inter-speaker variations are reduced by frequency and time normalisation, thereby bringing out common elements as well as individual global traits. The extent of F0 modulation of accent patterns are modelled as a function of the continuously scaled syllable and word prominence parameter Rs which we have introduced in earlier studies of stress and accentuation. Additional modifications with respect to the relative location of a word in a clause are introduced. Differences between speaker average and predicted data within a sentence are usually but not always small and will be systematically studied in order to derive rules for assimilation and syntactically motivated grouping. Our study contributes to detailed insights in the realisation of accent 1 and accent 2 patterns.
Bruce, G. Swedish Word Accents in Sentence Perspective. Lund: Gleerup, 1977.
Cite as: Fant, G., Kruckenberg, A. (2000) A prominence based model of Swedish intonation. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 2, 341-344, doi: 10.21437/ICSLP.2000-278
@inproceedings{fant00_icslp,
author={Gunnar Fant and Anita Kruckenberg},
title={{A prominence based model of Swedish intonation}},
year=2000,
booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
pages={vol. 2, 341-344},
doi={10.21437/ICSLP.2000-278}
}