F0 derivatives in the classification of meaningful tonal movements

Datos Bibliográficos

ID6006403
AutoresConstantijn Kaland (0000-0002-1813-5902, University of Cologne, autor de correspondencia)
Año2025
Volumen113
Páginas101454
Fecha de publicación2025-11-01
Peer ReviewedSí
Open AccessSí
TipoARTICLE
RevistaJournal of Phonetics (JOURNAL)
Identificadores de la revistaISSN: 0095-4470 • E-ISSN: 1095-8576
EditorialElsevier BV (PUBLISHER)
DOI10.1016/j.wocn.2025.101454
OpenAlexW4415889997
IdiomaEN
Referencias citadas76

F0 derivatives are informative for automatic classification of meaningful f0 movements. • Classification on the basis of f0 derivatives outperforms static or more direct measures of f0. • The third derivative (jerk) is particularly successful across multiple clusterings. Recent work applied cluster analysis on f0 contours in order to find ‘prototypical’ or ‘underlying’ categories as assumed in intonational phonology. However, it remains to be answered to what extent meaningful f0 variation can indeed be captured using automatic classification of surface realizations. Studies on f0 dynamics have suggested that derivatives (e.g., f0 velocity, acceleration and jerk) closely approximate the meaningful components of f0. The question answered in this study is to what extent f0 derivatives are more informative for cluster analysis than other metrics, such as the (time series) f0 contour they are derived from, a static measure representing it, or other acoustic measures such as intensity and duration. This is tested across two clustering techniques (hierarchical and k-medoids) for three different meaningful features expressed in Dutch noun phrases (of the type ‘blue sofa’): focus type (broad, narrow), focus position (adjective, noun) and phrase position (medial, final). Results show that derivatives are among the most informative acoustic measures, although the best performing cluster analyses are the ones based on multiple acoustic measures. Crucially, cluster analyses reveal that the different meaningful prosodic features each have their own characteristics in terms of acoustics and number of clusters

Basis (linear algebra) · Cluster (spacecraft) · Cluster analysis · Focus (optics) · Measure (data warehouse) · Pattern recognition (psychology) · Phrase · Position (finance) · Linguistic Variation and Morphology · Phonetics and Phonology Research · Voice and Speech Disorders