The ESCA Workshop on Speech Synthesis
September 25-28, 1990
In this paper we shall present an experiment which tries to automise the acquisition of prosodic knowledge by using a database. This base was compiled by analysing a corpus read by two speakers which was subsequently segmented and labelled by experts. Once the data was structured and the prosodic facts divided into a certain number of classes, a symbolic learning process was implemented on these sets. This enabled us to obtain rules which were logically expressed as regards the duration and the melody.
Bibliographic reference. Mortamet, L. / Emerard, Francoise / Miclet, Laurent (1990): "Attempting automatic prosodic knowledge acquisition using a database", In SSW1-1990, 209-214.