LiteASR: Efficient Automatic Speech Recognition with Low-Rank Approximation
This work addresses efficiency challenges in deploying large ASR models, offering a practical solution for resource-constrained applications, though it is incremental as it builds on existing compression techniques.
The paper tackles the computational bottleneck of ASR encoder models like Whisper by proposing LiteASR, a low-rank compression scheme that reduces encoder size by over 50% while maintaining or improving transcription accuracy compared to Whisper medium.
Modern automatic speech recognition (ASR) models, such as OpenAI's Whisper, rely on deep encoder-decoder architectures, and their encoders are a critical bottleneck for efficient deployment due to high computational intensity. We introduce LiteASR, a low-rank compression scheme for ASR encoders that significantly reduces inference costs while maintaining transcription accuracy. Our approach leverages the strong low-rank properties observed in intermediate activations: by applying principal component analysis (PCA) with a small calibration dataset, we approximate linear transformations with a chain of low-rank matrix multiplications, and further optimize self-attention to work in reduced dimensionality. Evaluation results show that our method can compress Whisper large-v3's encoder size by over 50%, matching Whisper medium's size with better transcription accuracy, thereby establishing a new Pareto frontier of accuracy and efficiency. The code of LiteASR is available at https://github.com/efeslab/LiteASR.