The Zero Resource Speech Challenge 2017

Ewan Dunbar (1, 2) , Xuan-Nga Cao (3, 2) , Juan Benjumea (1, 2) , Julien Karadayi (1, 2) , Mathieu Bernard (1, 2) , Laurent Besacier (4, 5) , Xavier Anguera (6) , Emmanuel Dupoux (3, 2)

1 EHESS - École des hautes études en sciences sociales
2 CoML - Apprentissage machine et développement cognitif
3 LSCP - Laboratoire de sciences cognitives et psycholinguistique
4 LIG - Laboratoire d'Informatique de Grenoble
5 GETALP - Groupe d’Étude en Traduction Automatique/Traitement Automatisé des Langues et de la Parole
6 Telefonica I+D - Telefonica Investigación y Desarrollo

We describe a new challenge aimed at discovering subword and word units from raw speech. This challenge is the followup to the Zero Resource Speech Challenge 2015. It aims at constructing systems that generalize across languages and adapt to new speakers. The design features and evaluation metrics of the challenge are presented and the results of seventeen models are discussed.

Domaines: Informatique et langage [cs.CL]

Ewan Dunbar, Xuan-Nga Cao, Juan Benjumea, Julien Karadayi, Mathieu Bernard, et al.. The Zero Resource Speech Challenge 2017. ASRU 2017, Dec 2017, Okinawa, Japan. ⟨hal-01687504⟩