Description du sujet de thèse

This PhD topic will take advantage of data collected in the sociological component of the ANR VOLI project, aiming to develop a methodology for the analysis and modeling of oral data for use in the social sciences and humanities (SSH) disciplines. Ultimately, it seeks to implement a research support tool capable of automatically identifying spoken regions or “areas of interest” wrt scientific needs in SSH within these data.
The proof of concept will rely on a corpus of interviews conducted in Spanish with micro-workers from Latin America, enriched with socio-demographic metadata collected via questionnaires. These oral corpora, used to support sociological hypotheses, will be analyzed through the lens of segmental and suprasegmental linguistic variation as well as automatic modeling, with the goal of detecting spoken regions potentially rich in analytical value for the SSH.
The identification of these “areas of interest” will be based on correlations between acoustic, prosodic, and expressive cues (such as intensity, emotional markers, etc.) and will be validated through several dimensions, including manual validation by SSH researchers and comparison with other corpora involving different professions and languages.
The ultimate goal is to propose an approach suitable for generalization and automation to assist researchers in the qualitative analysis of oral corpora, while offering substantial time savings and encouraging reflection on their interpretive strategies.

Activities:
The PhD work will focus on the following four main areas:
1. Statistical analysis and modeling of acoustic and prosodic variation in speech: (1) definition, extraction, and statistical modeling of acoustic and prosodic variation descriptors; (2) correlation with other linguistic levels (e.g., POS tagging).
2. Construction of variation patterns based on different variables (socio-demographic or derived from qualitative exploration of the corpus) and statistical validation (PCA, clustering, etc.).
3. Modeling expressivity: (1) extraction of acoustic measures related to emotion (activation, valence, intensity, etc.); (2) analysis of their relevance compared to manual annotations (qualitative labeling done by sociologists or linguists).
4. Automatic detection of “expressive” or “areas of interest” using classification, segmentation, or machine learning algorithms, and implementation of AI/ML models to automate the identification and comparison of relevant zones within oral corpora.
Expected Skills:
Linguistics and affective sciences: corpus linguistics, experimental phonetics, segmental and suprasegmental analysis of speech, expressivity modeling, etc. Knowledge of sociolinguistics applied to spoken variation is a plus.
Computing skills:
• Automatic speech processing (Python, Praat, OpenSMILE, etc.);
• Statistical modeling, classification, machine learning;
• Familiarity with linguistic annotation tools (e.g., ELAN, Praat).
Personal qualities:
• Interest in interdisciplinary research;
• Interest in oral corpora and human sciences;
• Ability to work collaboratively within multidisciplinary teams.

Contexte de travail

This PhD topic is part of the VOLI project (Voices from Online Labour: Inequalities in digital earning activities across countries), funded by the ANR for the 2024–2028 period and coordinated at LISN by Ioana Vasilescu. This interdisciplinary project innovatively combines hypotheses and methods from sociology, large-scale corpus linguistics, speech technologies, and artificial intelligence.

In addition to addressing the economic and social issues related to platform-based digital labor, VOLI focuses on linguistic variation in the spoken language of platform workers (referred to as “AI micro-workers”), by enriching the corpora with metadata from sociological surveys. At the same time, VOLI contributes to the development of new tools for the analysis of spoken language variation and the automatic processing of speech.

Le poste se situe dans un secteur relevant de la protection du potentiel scientifique et technique (PPST), et nécessite donc, conformément à la réglementation, que votre arrivée soit autorisée par l'autorité compétente du MESR.

Contraintes et risques

The selected candidate will be expected to actively participate in the activities of the LISN laboratory and the VOLI project. The position does not involve any particular risks.
To apply, please submit two documents: (1) a detailed CV and (2) a cover letter explaining your interest in and qualifications for this position.