More about deep learning in phonetics...

Emmanuel Ferragne

Associate Professor of Phonetics

Laboratoire de Phonétique et Phonologie

About me

I’ve been a Maître de Conférences (Associate Professor) in English phonetics at Université de Paris since 2009. January-June 2020 & 2021: full-time CNRS researcher at Laboratoire de Phonétique et Phonologie.

I initially studied translation, then moved to English studies where I developed a keen interest for linguistics and phonetics. I’ve always enjoyed the most technical aspects of my job: working with instruments (ultrasound imaging, EEG, EGG, etc.), programming and data analysis (Praat, R, Matlab, C#, Python, etc.). I’ve been doing this for 20 years so, although I’ve had no formal training, you could probably call me a data scientist.

Interests

{Socio|Neuro}-Phonetics
Forensic voice comparison
AI & Humanities
Metal Studies

Featured Publications

Labiodentals /r/ here to stay: Deep learning shows us why

Our results suggest that there is a recognisable difference between the lip postures for /r/ and /w/, which a convolutional neural network is able to detect with a very high degree of accuracy. Measurements of the lip area acquired using an artificial neural network suggest that /r/ indeed has a labiodental-like lip posture, thus providing a phonetic account for labiodentalisation. We finish with a discussion of the methodological implications of using deep learning for future analyses of phonetic data.

Hannah King, Emmanuel Ferragne

Anglophonia, 2021

Code Full text

Recent Publications

More Publications

Neuroplasticity in the phonological system: The PMN and the N400 as markers for the perception of non-native phonemic contrasts by late second language learners

Second language (L2) learners frequently encounter persistent difficulty in perceiving certain non-native sound contrasts, i.e., a …

Karin Heidlmayr, Emmanuel Ferragne, Frédéric Isel

DOI

The production and perception of derived phonological contrasts in selected varieties of English

My goal is to examine acoustic and perceptual data from various locations in the British Isles – Enniskillen (Ulster), Glasgow, and …

Emmanuel Ferragne

book

Paramètres acoustiques et phonétiques dans la parole parkinsonienne avant et après traitement LSVT LOUD®

Notre recherche examine l’effet du Lee Silverman Voice Treatment (LSVT LOUD®) sur l’aire et la position de l’espace vocalique, la …

Maëlle Le Cerf, Emmanuel Ferragne

PDF

Informations segmentales pour la caractérisation phonétique du locuteur : variabilité inter- et intra-locuteurs

Nous avons effectué une classification automatique de 44 locuteurs à partir de réseaux de neurones convolutifs (CNN) sur la base de …

Cédric Gendrot, Emmanuel Ferragne, Thomas Pellegrini

PDF

Caractérisation du locuteur par CNN à l’aide des contours d’intensité et d’intonation : comparaison avec le spectrogramme

Dans ce travail nous avons recours aux variations de f0 et d’intensité de 44 locuteurs francophones à partir de séquences de 4 secondes …

Gabriele Chignoli, Cédric Gendrot, Emmanuel Ferragne

PDF

Loose lips and tongue tips: The central role of the /r/-typical labial gesture in Anglo-English

Hannah King, Emmanuel Ferragne

PDF DOI

The perception of swear words by French learners of English: an experiment involving electrodermal activity

Hearing swear words or taboo words causes us discomfort. Research suggests that emotional responses caused by such words are stronger …

Anastasija Rastovic, Maud Pélissier, Emmanuel Ferragne

DOI Full text

The contribution of lip protrusion to Anglo-English /r/: Evidence from hyper- and non-hyperarticulated speech

Hannah King, Emmanuel Ferragne

PDF DOI

A combined forced-attention dichotic listening - Go/Nogo task to assess response inhibition and interference suppression: An auditory event-related potential investigation

OBJECTIVE: Interference suppression and response inhibition are distinct effortful inhibitory processes. Yet they rely on partly …

Nathalie Bedoin, Raphaëlle Abadie, Jennifer Krzonowski, Emmanuel Ferragne, Agathe Marcastel

DOI

Deep learning and voice comparison: phonetically-motivated vs. automatically-learned features

Broadband spectrograms of French vowels /ɑ̃/, /a/, /ɛ/, /e/, /i/, /ə/, and /ɔ/ extracted from radio broadcast corpora were used to …

Cédric Gendrot, Emmanuel Ferragne, Thomas Pellegrini

PDF Project

Featured Talks

Du groupe à l'individu, du corpus à l'expérimentation, du spectrogramme au deep learning pour la phonétique

Jun 25, 2021 2:00 PM

Soutenance Habilitation à Diriger des Recherches

Since scientists’ individual epistemological preferences infallibly shape the output of their research, this thesis starts with a presentation of the author’s position with respect to a number of methodological issues pertaining to the field of contemporary phonetics…