Juliette Millet

Ancien doctorant

Statut : Doctorante

Adresse :

LLF, CNRS – UMR 7110
Université Paris Diderot-Paris 7
Case 7031 – 5, rue Thomas Mann,
75205 Paris cedex 13

E-mail : whyvrggr.zvyyrg@pev-cnevf.bet

CV : Juliette Millet_fr.pdf (28 Ko)

Site Web : https://jamju.github.io/


Titre : Modelling non-native speech perception using automatic speech processing

Directeur :

Date de soutenance : 2022-05-16

Inscription : 2018 à Paris Cité

Jury :

  • Laurent Besacier,
  • Ioana Chitoran,
  • Ewan Dunbar,
  • Emmanuel Dupoux,
  • Laura Gwilliams,
  • Odette Scharenborg,
  • Jean-Luc Schwartz

Résumé :

My thesis deals with modeling foreign language speech perception and learning. Large-scale computational models of speech perception are never tested for other languages than the one they learned. Hence no real modeling has been done concerning the difficulty people have with other languages than their own. Doing so could be interesting for improving the models, making them more 'human-like', and for enabling us to predict what difficulties humans need to focus on if they learn a new language. The project has three goals: predicting difficulties in foreign speech perception, predicting outcomes in foreign language learning and modeling intelligibility in a foreign language.