Closing the Speech-Text Gap with Limited Audio for Effective Domain Adaptation in LLM-Based ASR
This addresses domain adaptation for ASR systems, offering a more efficient approach by reducing the need for extensive paired speech-text data, though it is incremental as it builds on existing LLM-based ASR methods.
The paper tackled the modality gap in LLM-based ASR adaptation by using limited speech data, finding that mixed batching with only 10% of target-domain speech (less than 4 hours) achieves word error rates comparable to or better than full-dataset fine-tuning.
Conventional end-to-end automatic speech recognition (ASR) systems rely on paired speech-text data for domain adaptation. Recent LLM-based ASR architectures connect a speech encoder to a large language model via a projection module, enabling adaptation with text-only data. However, this introduces a modality gap, as the LLM is not exposed to the noisy representations produced by the speech projector. We investigate whether small amounts of speech can mitigate this mismatch. We compare three strategies: text-only adaptation, paired speech-text adaptation, and mixed batching (MB), which combines both. Experiments in in-domain and out-of-domain settings show that even limited speech consistently improves performance. Notably, MB using only 10% of the target-domain (less than 4 hours) speech achieves word error rates comparable to, or better than, conventional ASR fine-tuning with the full dataset, indicating that small amounts of speech provide a strong modality-alignment signal.