SSHR: Leveraging Self-supervised Hierarchical Representations for Multilingual Automatic Speech Recognition

TL;DR

This paper introduces SSHR, a novel approach that leverages hierarchical representations in self-supervised models to improve multilingual automatic speech recognition, achieving state-of-the-art results.

Contribution

The study proposes a method to exploit different layer representations in SSL models for better multilingual ASR, including a novel hierarchical extraction and guidance mechanism.

Findings

Middle layers contain language-related info

High layers encode content-related info

Method achieves state-of-the-art performance on benchmarks

Abstract

Multilingual automatic speech recognition (ASR) systems have garnered attention for their potential to extend language coverage globally. While self-supervised learning (SSL) models, like MMS, have demonstrated their effectiveness in multilingual ASR, it is worth noting that various layers' representations potentially contain distinct information that has not been fully leveraged. In this study, we propose a novel method that leverages self-supervised hierarchical representations (SSHR) to fine-tune the MMS model. We first analyze the different layers of MMS and show that the middle layers capture language-related information, and the high layers encode content-related information, which gradually decreases in the final layers. Then, we extract a language-related frame from correlated middle layers and guide specific language extraction through self-attention mechanisms. Additionally, we steer the model toward acquiring more content-related information in the final layers using our proposed Cross-CTC. We evaluate SSHR on two multilingual datasets, Common Voice and ML-SUPERB, and the experimental results demonstrate that our method achieves state-of-the-art performance.