Fast Development of ASR in African Languages using Self Supervised Speech Representation Learning

TL;DR

This paper demonstrates that self-supervised speech representation learning enables effective ASR development for low-resource African languages using minimal transcribed data, highlighting the importance of pre-training on large raw speech datasets.

Contribution

It introduces a practical approach for developing ASR systems for African languages with limited labeled data, leveraging self-supervised learning and pre-training techniques.

Findings

Pre-training on large raw speech datasets improves ASR performance.

Effective ASR systems can be built with only 1 hour of transcribed speech.

Self-supervised learning is crucial in low-resource language settings.

Abstract

This paper describes the results of an informal collaboration launched during the African Master of Machine Intelligence (AMMI) in June 2020. After a series of lectures and labs on speech data collection using mobile applications and on self-supervised representation learning from speech, a small group of students and the lecturer continued working on automatic speech recognition (ASR) project for three languages: Wolof, Ga, and Somali. This paper describes how data was collected and ASR systems developed with a small amount (1h) of transcribed speech as training data. In these low resource conditions, pre-training a model on large amounts of raw speech was fundamental for the efficiency of ASR systems developed.