Curriculum optimization for low-resource speech recognition

TL;DR

This paper introduces an automated curriculum learning method that optimizes training example sequences for low-resource speech recognition, significantly improving Word Error Rate performance by up to 33%.

Contribution

It presents a novel difficulty measure called compression ratio and an automated curriculum approach tailored for low-resource speech recognition tasks.

Findings

Up to 33% relative WER reduction over baseline

Effective use of compression ratio as difficulty measure

Improved training efficiency for low-resource speech data

Abstract

Modern end-to-end speech recognition models show astonishing results in transcribing audio signals into written text. However, conventional data feeding pipelines may be sub-optimal for low-resource speech recognition, which still remains a challenging task. We propose an automated curriculum learning approach to optimize the sequence of training examples based on both the progress of the model while training and prior knowledge about the difficulty of the training examples. We introduce a new difficulty measure called compression ratio that can be used as a scoring function for raw audio in various noise conditions. The proposed method improves speech recognition Word Error Rate performance by up to 33% relative over the baseline system