Can I read SAML: Speaker Adaptive Mixture of LoRA Experts for End-to-End ASR on EtoBox?

SAML: Speaker Adaptive Mixture of LoRA Experts for End-to-End ASR by Zhao, Qiuming; Sun, Guangzhi; Zhang, Chao; Xu, Mingxing; Zheng, Thomas Fang is a scholarly article available to read on EtoBox.

What is SAML: Speaker Adaptive Mixture of LoRA Experts for End-to-End ASR about?

Mixture-of-experts (MoE) models have achieved excellent results in many tasks. However, conventional MoE models are often very large, making them challenging to deploy on resource-constrained edge devices. In this paper, we propose a novel speaker adaptive mixture of LoRA experts (SAML) approach, which uses low-rank adaptation (LoRA) modules as experts to reduce the number of trainable parameters in MoE. Specifically, SAML is applied to the quantised and personalised end-to-end automatic speech recognition models, which combines test-time speaker adaptation to improve the performance of heavily compressed models in speaker-specific scenarios. Experiments have been performed on the LibriSpeech and the TED-LIUM 3 corpora. Remarkably, with a 7x reduction in model size, 29.1% and 31.1% relative word error rate reductions were achieved on the quantised Whisper model and Conformer-based attention-based encoder-decoder ASR model respectively, comparing to the original full precision models.

Author

Zhao, Qiuming; Sun, Guangzhi; Zhang, Chao; Xu, Mingxing; Zheng, Thomas Fang

Published

2024

Language

EN