Classification of Swedish dialects using a hierarchical prosodic analysis

Marcin Włodarczak, Juraj Šimko, Antti Suni, Martti Vainio


The present study investigates dialectal variation of Swedish word accents by means of wavelet-based analysis of F0 and energy. The analysis yields a measure of prosodic similarity between dialects expressed in terms of mutual perplexity of unigram models trained on derivatives of the wavelet-decomposed input features. A comparison of models trained on energy, F0 and a combination of both features indicates that the energy+F0 model reaches the highest classification accuracy, in line with the existing descriptions of tonal dialects in terms of the number and timing of pitch peaks with respect to the stressed syllable. At the same time, prosodic similarity between geographically close but typologically distinct dialects suggests an interaction between the traditional distinction between type-1 and type-2 dialects and regional variation, giving rise to northern and southern type-2 dialects (with little difference between 2A and 2B subtypes), and a parallel distinction between 1A and 1B varieties.


 DOI: 10.21437/SpeechProsody.2018-62

Cite as: Włodarczak, M., Šimko, J., Suni, A., Vainio, M. (2018) Classification of Swedish dialects using a hierarchical prosodic analysis. Proc. 9th International Conference on Speech Prosody 2018, 304-308, DOI: 10.21437/SpeechProsody.2018-62.


@inproceedings{Włodarczak2018,
  author={Marcin Włodarczak and Juraj Šimko and Antti Suni and Martti Vainio},
  title={Classification of Swedish dialects using a hierarchical prosodic analysis},
  year=2018,
  booktitle={Proc. 9th International Conference on Speech Prosody 2018},
  pages={304--308},
  doi={10.21437/SpeechProsody.2018-62},
  url={http://dx.doi.org/10.21437/SpeechProsody.2018-62}
}