Logo image
Back
Augmenting Conformers With Structured State-Space Sequence Models For Online Speech Recognition
Conference proceeding   Open access

Augmenting Conformers With Structured State-Space Sequence Models For Online Speech Recognition

Haozhe Shan, Albert Gu, Zhong Meng, Weiran Wang, Krzysztof Choromanski and Tara Sainath
ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp.12221-12225
04/14/2024
DOI: 10.1109/ICASSP48485.2024.10445950
url
https://doi.org/10.1109/ICASSP48485.2024.10445950View
Published (Version of record) Open Access

Abstract

Online speech recognition, where the model only accesses context to the left, is an important and challenging use case for ASR systems. In this work, we investigate augmenting neural encoders for online ASR by incorporating structured state-space sequence models (S4), a family of models that provide a parameter-efficient way of accessing arbitrarily long left context. We performed systematic ablation studies to compare variants of S4 models and propose two novel approaches that combine them with convolutions. We found that the most effective design is to stack a small S4 using real-valued recurrent weights with a local convolution, allowing them to work complementarily. Our best model achieves WERs of 4.01%/8.53% on test sets from Librispeech, outperforming Conformers with extensively tuned convolution.
 Acoustics causal model Conformer Context modeling Convolution Online ASR Speech processing Speech recognition state-space model Systematics

Details

Metrics

15 Record Views
Logo image