Improving Joint Speech-Text Representations Without Alignment
This work addresses a bottleneck in speech recognition by simplifying alignment in joint encoders, though it is incremental as it builds on existing methods.
The paper tackles the problem of sequence-length mismatch in joint speech-text encoders by proposing a consistency loss that forgives length differences without explicit alignment, improving word error rate in both monolingual and multilingual systems.
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.