Guided contrastive self-supervised pre-training for automatic speech recognition

TL;DR

This paper introduces Guided Contrastive Predictive Coding (GCPC), a novel pre-training method for automatic speech recognition that incorporates prior knowledge to improve model performance across multiple languages.

Contribution

The paper proposes GCPC, a new contrastive learning approach that injects prior knowledge during pre-training to enhance speech recognition accuracy.

Findings

GCPC outperforms CPC pre-training on all three datasets.

Reduces Word Error Rate (WER) by up to 15.43%.

Demonstrates effectiveness across German, French, and English ASR tasks.

Abstract

Contrastive Predictive Coding (CPC) is a representation learning method that maximizes the mutual information between intermediate latent representations and the output of a given model. It can be used to effectively initialize the encoder of an Automatic Speech Recognition (ASR) model. We present a novel modification of CPC called Guided Contrastive Predictive Coding (GCPC). Our proposed method maximizes the mutual information between representations from a prior-knowledge model and the output of the model being pre-trained, allowing prior knowledge injection during pre-training. We validate our method on 3 ASR tasks: German, French and English. Our method outperforms CPC pre-training on all three datasets, reducing the Word Error Rate (WER) by 4.44%, 6.55% and 15.43% relative on the German, French and English (Librispeech) tasks respectively, compared to training from scratch, while CPC pre-training only brings 2.96%, 1.01% and 14.39% relative WER reduction respectively.