English

How Much Context Does My Attention-Based ASR System Need?

Computation and Language 2024-06-18 v2 Sound Audio and Speech Processing

Abstract

For the task of speech recognition, the use of more than 30 seconds of acoustic context during training is uncommon and under-investigated in literature. In this work, we conduct an empirical study on the effect of scaling the sequence length used to train/evaluate (dense-attention-based) acoustic models on speech recognition performance. For these experiments, a dataset of roughly 100,000 pseudo-labelled Spotify podcasts is used, with context lengths of 5 seconds to 1 hour being explored. Zero-shot evaluations are presented on the long-format datasets: Earnings-22, Tedlium and Rev16. Results demonstrate a benefit from training with up to 21.8 minutes of acoustic context, showing up to a 14.5\% relative improvement from a baseline trained with 10 seconds of context. We find that the model's width/depth, positional encoding scheme and number of attention heads impact its ability to use longer contexts.

Keywords

Cite

@article{arxiv.2310.15672,
  title  = {How Much Context Does My Attention-Based ASR System Need?},
  author = {Robert Flynn and Anton Ragni},
  journal= {arXiv preprint arXiv:2310.15672},
  year   = {2024}
}

Comments

Accepted at Interspeech 2024

R2 v1 2026-06-28T13:00:01.798Z