Skip to Main content Skip to Navigation
Master Thesis

Révision de données phonétiques et acoustiques & transformation de dépendances de surface en dépendances profondes de Stanford

Abstract : This paper describes the realization of two different tasks as part of two projects. The first one consists of the manual checking of phonetic and acoustic data that will form the phoneme base of a speech synthesizer. After a brief study of the speech synthesis model, we review and classify the phonetic errors generated by the grapheme-to-phoneme tool, comparing the possible ways of correcting them (either automatically or manually). Then, we distinguish categories of errors implying phonemic barriers shifting after sound segmentation, which is aimed at matching the sound with the phonemes of the transcriptions. In the second section, we describe the implementation of a module of rules for a graph transformation model to convert parses from shallow typed dependencies into the deeper Stanford typed dependencies, which we adjusted to the French language. Finally, we perform an brief evaluation of the module, and suggest some improvements in order to increase processing speed.
Document type :
Master Thesis
Complete list of metadata

Cited literature [11 references]  Display  Hide  Download

https://dumas.ccsd.cnrs.fr/dumas-00879928
Contributor : Uga - Bulles <>
Submitted on : Tuesday, November 5, 2013 - 10:09:55 AM
Last modification on : Tuesday, May 11, 2021 - 11:36:17 AM
Long-term archiving on: : Friday, April 7, 2017 - 9:27:51 PM

Identifiers

  • HAL Id : dumas-00879928, version 1

Citation

Cécile Robin. Révision de données phonétiques et acoustiques & transformation de dépendances de surface en dépendances profondes de Stanford. Sciences de l'Homme et Société. 2013. ⟨dumas-00879928⟩

Share

Metrics

Record views

239

Files downloads

504