Listen, Adapt, Better WER: Source-free Single-utterance Test-time Adaptation for Automatic Speech Recognition
This addresses the need for robust ASR in real-world scenarios where test data distributions vary, offering a practical adaptation method without batch delays.
The paper tackles the problem of performance regression in automatic speech recognition (ASR) when test data comes from different distributions, proposing a single-utterance test-time adaptation framework that improves word error rates on out-of-domain and in-domain test samples without accessing source data.
Although deep learning-based end-to-end Automatic Speech Recognition (ASR) has shown remarkable performance in recent years, it suffers severe performance regression on test samples drawn from different data distributions. Test-time Adaptation (TTA), previously explored in the computer vision area, aims to adapt the model trained on source domains to yield better predictions for test samples, often out-of-domain, without accessing the source data. Here, we propose the Single-Utterance Test-time Adaptation (SUTA) framework for ASR, which is the first TTA study on ASR to our best knowledge. The single-utterance TTA is a more realistic setting that does not assume test data are sampled from identical distribution and does not delay on-demand inference due to pre-collection for the batch of adaptation data. SUTA consists of unsupervised objectives with an efficient adaptation strategy. Empirical results demonstrate that SUTA effectively improves the performance of the source ASR model evaluated on multiple out-of-domain target corpora and in-domain test samples.