Spotting Prosodic Boundaries in Continuous Speech in French

TL;DR

This paper presents a system for automatically detecting prosodic boundaries in French speech using a neural network trained on a large annotated corpus, validated by expert and non-expert annotations.

Contribution

It introduces a new large French speech corpus with prosodic markings and a neural network approach for boundary detection using multiple acoustic features.

Findings

Successful automatic boundary spotting validated by expert annotations

Effective use of F0, vowels, and pseudo-syllable durations as features

Large annotated corpus enables robust system training

Abstract

A radio speech corpus of 9mn has been prosodically marked by a phonetician expert, and non expert listeners. this corpus is large enough to train and test an automatic boundary spotting system, namely a time delay neural network fed with F0 values, vowels and pseudo-syllable durations. Results validate both prosodic marking and automatic spotting of prosodic events.