token2vec: A Joint Self-Supervised Pre-training Framework Using Unpaired Speech and Text
This work addresses the challenge of integrating complementary speech and text data for improved speech processing tasks, representing an incremental advance in multimodal pre-training.
The paper tackles the problem of joint self-supervised pre-training for unpaired speech and text by proposing token2vec, which discretizes speech and aligns text phonemes to address modality and length mismatches, resulting in up to 17.7% relative WER reduction compared to speech-only baselines.
Self-supervised pre-training has been successful in both text and speech processing. Speech and text offer different but complementary information. The question is whether we are able to perform a speech-text joint pre-training on unpaired speech and text. In this paper, we take the idea of self-supervised pre-training one step further and propose token2vec, a novel joint pre-training framework for unpaired speech and text based on discrete representations of speech. Firstly, due to the distinct characteristics between speech and text modalities, where speech is continuous while text is discrete, we first discretize speech into a sequence of discrete speech tokens to solve the modality mismatch problem. Secondly, to solve the length mismatch problem, where the speech sequence is usually much longer than text sequence, we convert the words of text into phoneme sequences and randomly repeat each phoneme in the sequences. Finally, we feed the discrete speech and text tokens into a modality-agnostic Transformer encoder and pre-train with token-level masking language modeling (tMLM). Experiments show that token2vec is significantly superior to various speech-only pre-training baselines, with up to 17.7% relative WER reduction. Token2vec model is also validated on a non-ASR task, i.e., spoken intent classification, and shows good transferability.