English

WhisBERT: Multimodal Text-Audio Language Modeling on 100M Words

Computation and Language 2023-12-08 v2 Artificial Intelligence

Abstract

Training on multiple modalities of input can augment the capabilities of a language model. Here, we ask whether such a training regime can improve the quality and efficiency of these systems as well. We focus on text--audio and introduce Whisbert, which is inspired by the text--image approach of FLAVA (Singh et al., 2022). In accordance with Babylm guidelines (Warstadt et al., 2023), we pretrain Whisbert on a dataset comprising only 100 million words plus their corresponding speech from the word-aligned version of the People's Speech dataset (Galvez et al., 2021). To assess the impact of multimodality, we compare versions of the model that are trained on text only and on both audio and text simultaneously. We find that while Whisbert is able to perform well on multimodal masked modeling and surpasses the Babylm baselines in most benchmark tasks, it struggles to optimize its complex objective and outperform its text-only Whisbert baseline.

Keywords

Cite

@article{arxiv.2312.02931,
  title  = {WhisBERT: Multimodal Text-Audio Language Modeling on 100M Words},
  author = {Lukas Wolf and Greta Tuckute and Klemen Kotar and Eghbal Hosseini and Tamar Regev and Ethan Wilcox and Alex Warstadt},
  journal= {arXiv preprint arXiv:2312.02931},
  year   = {2023}
}

Comments

Published at the BabyLM Challenge, a shared task co-sponsored by CMCL 2023 and CoNLL 2023, hosted by EMNLP 2023

R2 v1 2026-06-28T13:41:55.696Z