Improving Joint Speech-Text Representations Without Alignment

TL;DR

This paper proposes a method to improve joint speech-text representations by ignoring sequence length mismatches, leading to better speech recognition performance without complex alignment techniques.

Contribution

It introduces a consistency loss that allows joint speech-text encoders to handle sequence length differences naturally, enhancing downstream WER results.

Findings

Improved WER in monolingual systems

Enhanced multilingual speech recognition performance

Sequence length mismatch can be effectively ignored

Abstract

The last year has seen astonishing progress in text-prompted image generation premised on the idea of a cross-modal representation space in which the text and image domains are represented jointly. In ASR, this idea has found application as joint speech-text encoders that can scale to the capacities of very large parameter models by being trained on both unpaired speech and text. While these methods show promise, they have required special treatment of the sequence-length mismatch inherent in speech and text, either by up-sampling heuristics or an explicit alignment model. In this work, we offer evidence that joint speech-text encoders naturally achieve consistent representations across modalities by disregarding sequence length, and argue that consistency losses could forgive length differences and simply assume the best alignment. We show that such a loss improves downstream WER in both a large-parameter monolingual and multilingual system.