Generative Spoken Language Model based on continuous word-sized audio tokens

Robin Jonathan Algayres, Yossi Adi, Tu Anh Nguyen, Jade Copet, Gabriel Synnaeve, Benoît Sagot, Emmanuel Dupoux

Submission Type: Regular Long Paper

Submission Track: Speech and Multimodality

Submission Track 2: Language Modeling and Analysis of Language Models

Keywords: spoken language models, speech generation, zerospeech, textless nlp

TL;DR: A spoken language models that trained on word-size continuous acoustic embeddings

Abstract: In NLP, text language models based on words or subwords are known to outperform their character-based counterparts. Yet, in the speech community, the standard input of spoken LMs are 20ms or 40ms-long discrete units (shorter than a phoneme). Taking inspiration from word-based LM, we introduce a Generative Spoken Language Model (GSLM) based on word-size continuous-valued audio tokens that can generate diverse and expressive language output. This is obtained by replacing lookup table for lexical types with a Lexical Embedding function, the cross entropy loss by a contrastive loss, and multinomial sampling by k-NN sampling. The resulting model is the first generative language model based on word-size continuous tokens. Its performance is on par with discrete unit GSLMs regarding generation quality as measured by automatic metrics and subjective human judgements. Moreover, it is five times more memory efficient thanks to its large 200ms units. In addition, the embeddings before and after the Lexical Embedder are phonetically and semantically interpretable.

Submission Number: 110