Self-supervised language learning from raw audio: Lessons from the Zero Resource Speech Challenge

TL;DR

This paper reviews the Zero Resource Speech Challenge series, which advances self-supervised learning from raw audio for speech processing by defining tasks, metrics, and benchmarks, highlighting progress and future challenges.

Contribution

It provides an overview of six editions of the challenge series, summarizing lessons learned and identifying key areas for future research in zero-resource speech processing.

Findings

Progress in acoustic unit discovery and spoken term discovery

Development of benchmarks and metrics for model comparison

Identification of challenges and future directions in zero-resource speech learning

Abstract

Recent progress in self-supervised or unsupervised machine learning has opened the possibility of building a full speech processing system from raw audio without using any textual representations or expert labels such as phonemes, dictionaries or parse trees. The contribution of the Zero Resource Speech Challenge series since 2015 has been to break down this long-term objective into four well-defined tasks -- Acoustic Unit Discovery, Spoken Term Discovery, Discrete Resynthesis, and Spoken Language Modeling -- and introduce associated metrics and benchmarks enabling model comparison and cumulative progress. We present an overview of the six editions of this challenge series since 2015, discuss the lessons learned, and outline the areas which need more work or give puzzling results.