The Zero Resource Speech Challenge 2017

TL;DR

The paper introduces the 2017 Zero Resource Speech Challenge, focusing on unsupervised discovery of linguistic units from raw speech across languages and speakers, with evaluation of multiple models.

Contribution

It presents a new challenge framework for zero-resource speech processing, extending previous work to include cross-language and speaker adaptation capabilities.

Findings

Seventeen models evaluated on the challenge.

Models show varying success in unsupervised speech unit discovery.

Framework facilitates benchmarking for zero-resource speech systems.

Abstract

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.