Hot research topics
I’ll supervise the following projects for 2025-2026. For more info, please send me an email.
TOPIC 1: One speaker, two languages, two voices: forensic phonetics and bilingualism
Our everyday experience suggests that when a speaker switches languages, his or her voice changes too. This effect is generally attributable to two phenomena: first, the phonemes of the two languages require specific articulations leading to different “average configurations” of the vocal tract, known as articulatory settings. And in addition to these language-specific articulatory settings, the quality as well as the frequency of vocal fold vibration may be affected. For instance, a bilingual speaker may speak with a higher-pitched voice in his or her native language. Very few research articles have examined the impact of such changes on automatic speaker recognition systems or on people’s perceptual ability to recognize someone’s voice. The aim of this project, in collaboration with the forensic department of the French Police Nationale, is to precisely measure how code switching affects the performance of computers or ear witnesses in a speaker recognition task, and to pinpoint the phonetic parameters that underlie these performance fluctuations.
TOPIC 2: Medical imaging for the acquisition of phonology in a foreign language
It has been repeatedly suggested in the literature that when learners visualize their organs of speech, their pronunciation of a foreign language improves, potentially without the need for a teacher to formulate explicit phonetic instruction. The question therefore arises as to whether the state-of-the art techniques we have access to in our lab – ultrasound tongue imaging, electromagnetic articulography, and magnetic resonance imaging – can help French learners of English learn sounds that are known to be particularly difficult for them (some diphthongs, /l/, /r/, etc.). It is also relevant to ask which of these technologies offers optimal efficiency, and whether they can be used in the classroom or, at least, provide useful teaching material. In collaboration with the Groupe Hospitalier Universitaire (GHU) Paris, this project in experimental phonetics explores whether advanced medical imaging techniques can be repurposed for pronunciation teaching and research in second language acquisition.
TOPIC 3: Automatic accent identification with AI
It is now well-established that speaking English with a foreign accent may have undesirable consequences. To take but one example from the scientific literature: doctors who have a foreign accent are perceived as less reliable. And this negative effect holds true for native speakers with stigmatized accents (e.g., in the UK, Birmingham). As teachers of English, it is our duty to help our students make the best choices for their careers; and choosing an accent (even though the choice may not be explicit) when you are not a native speaker is one of them. Therefore, being able to analyze students’ oral productions and accents is of utmost importance for teachers and learners of foreign languages. However, doing this auditorily is impracticable with large cohorts of students. Fortunately recent AI-based models can perform this task for us. The project, relying on cutting-edge deep learning technology, aims to discover patterns in the pronunciation of French learners of English, and to automatically locate the phonemes in the audio signal that contribute the most to the perception of an accent (American, British, or French).
