MS-HuBERT: Mitigating Pre-training and Inference Mismatch in Masked Language Modelling methods for learning Speech Representations

TL;DR

MS-HuBERT introduces a novel approach to mitigate pre-training and inference mismatch in HuBERT, enhancing speech representations and improving ASR performance by 5% on Librispeech benchmark.

Contribution

The paper proposes the Swap method and Multicluster masked prediction loss to improve HuBERT's pre-training and inference alignment, leading to better speech representations.

Findings

MS-HuBERT outperforms vanilla HuBERT by 5% on Librispeech.

Learned embeddings encode essential information for ASR.

Proposed methods improve model capacity utilization.

Abstract

In recent years, self-supervised pre-training methods have gained significant traction in learning high-level information from raw speech. Among these methods, HuBERT has demonstrated SOTA performance in automatic speech recognition (ASR). However, HuBERT's performance lags behind data2vec due to disparities in pre-training strategies. In this paper, we propose (i) a Swap method to address pre-training and inference mismatch observed in HuBERT and (ii) incorporates Multicluster masked prediction loss for more effective utilization of the models capacity. The resulting method is, MS-HuBERT, an end-to-end self-supervised pre-training method for learning robust speech representations. It beats vanilla HuBERT on the ASR Librispeech benchmark on average by a 5% margin when evaluated on different finetuning splits. Additionally, we demonstrate that the learned embeddings obtained during pre-training encode essential information for improving performance of content based tasks such as ASR.