hi, I'm

Thibault Bañeras-Roux

what a speech recognizer heard: Thibault Roux [0.98]

I'm a postdoc at Idiap Research Institute, working on speech recognition and language models. I'm curious about meaning representation and how humans perceive and experience machine errors. Feel free to open discussion about low-resources languages, AI for good, unsupervised training or tech dangers.

Thibault smiling, with a forest behind him
TR

What I've done

Papers, dataset & toolkit.

Google Scholar

My papers

Evaluating speech recognition, how people perceive its errors, and metrics that go beyond Word Error Rate.

Open my profile
Hugging Face dataset

HATS-fr

An open French corpus for studying how humans perceive automatic speech recognition (ASR) errors, used to build evaluation metrics beyond Word Error Rate.

Open the dataset
GitHub dataset

HATS-en

An open English corpus for studying how humans perceive automatic speech recognition (ASR), extending HATS-fr.

Open the dataset
EU project toolkit

iRead4Skills: classifier and writing assistant

An open-access intelligent system that evaluates how complex a text is.

Open the toolkit

Where I've been

Marseille, Uppsala, Nantes, Avignon, Louvain-la-Neuve, Martigny.

2025 to now

Postdoc in NLP and speech processing

Idiap Research Institute

Working on LLM-based ASR (Automatic Speech Recognition) for low-resource domain adaptation.

Martigny, Switzerland
2024 to 2025

Postdoc in NLP

UCLouvain - CENTAL

In the context of the EU iRead4Skills project, I built text complexity tools that help adults strengthen their reading skills by suggesting texts at the right level for each person. Maintained and contributed to FABRA, a tool that extracts linguistic features from text. Supervised by Thomas François

Louvain-la-Neuve, Belgium
2021 to 2024

PhD in NLP and speech processing

Nantes University - LS2N and Avignon University - LIA

Analysis and Understanding of the Evaluation of Automatic Speech Recognition Systems: Towards Metrics Integrating Human Perception. I studied how people perceive ASR errors and metrics beyond Word Error Rate across classical and end-to-end systems. Part of the ANR DIETS project. Supervised by Richard Dufour, Jane Wottawa and Mickaël Rouvier.

Nantes and Avignon, France
2021

Internship in NLP

Aix-Marseille University - LIS, NLP team (TALEP)

Evaluated monolingual and cross-lingual embeddings, with a visualization method to see part-of-speech clusters, keeping low-resource languages in mind. Supervised by Carlos Ramisch and Manon Scholivet.

Marseille, France
2020 to 2021

Master 2nd year, Artificial Intelligence and Machine Learning

Aix-Marseille University

Marseille, France
2019 to 2020

Master 1st year, Artificial Intelligence

Uppsala University

Uppsala, Sweden
2019

Internship in Computer Science

LPL

Enriched the non-verbal behavior of an autonomous virtual patient to make human-machine interaction in virtual reality more believable. Supervised by Magalie Ochs and Philippe Blache.

Aix-en-Provence, France
2016 to 2019

Bachelor in Computer Science

Aix-Marseille University

Marseille, France

Teaching

2021 to 2024

Business Information Systems

2022 to 2024

Mobile Apps

2021 to 2023

Computer Structure

2021 to 2022

Operating Systems

2021 to 2022