About me

Hi there ! I am a PhD student based in Paris. My research focuses on learning language from child-like data ! By recording what stimuli children receive (speech, vision, etc.), we can build computational models behaving in a child-like manner. These models can be examined to predict developmental delays and learn more about what is useful for acquiring language. We can also exhibit failure cases where the child is better than the machine at doing its job in order to try to fix the machine.

Selected list of publications

A complete list of publications can be found here

An open-source voice type classifier for child-centered daylong recordings
Marvin Lavechin, Ruben Bousbib, Hervé Bredin, Emmanuel Dupoux, Alejandrina Cristia
ArXiV GitHub

End-to-end Domain-Adversarial Voice Activity Detection
Marvin Lavechin, Marie-Philippe Gill, Ruben Bousbib, Hervé Bredin, Leibny Paola Garcia-Perera
ArXiV GitHub

A thorough evaluation of the Language Environment Analysis (LENA) system
Alejandrina Cristia, Marvin Lavechin, Camila Scaff, Melanie Soderstrom, Caroline Rowland, Okko Räsänen, John Bunce, Elika Bergelson
ResearchGate GitHub

ALICE: An open-source tool for automatic measurement of phoneme, syllable, and word counts from child-centered daylong recordings
Okko Räsänen, Shreyas Seshadri, Marvin Lavechin, Alejandrina Cristia, Marisa Casillas
PsyArXiV GitHub