French Prominence: a Probabilistic Framework
Résumé
Identification of prosodic phenomena is of first importance inprosodic analysis and modeling. In this paper, we introduce anew method for automatic prosodic phenomena labelling. The au-thors set their approach of prosodic phenomena in the frameworkof prominence. The proposed method for automatic prominencelabelling is based on well-known machine learning techniques ina three step procedure: i) a feature extraction step in which wepropose a framework for systematic and multi-level speech acousticfeature extraction, ii) a feature selection step for identifying the morerelevant prominence acoustic correlates, and iii) a modelling step inwhich a gaussian mixture model is used for predicting prominence.This model shows robust performance on read speech (84%).
Domaines
LinguistiqueOrigine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...