ASR low‑resource language training

[libby-97]

I’d like to inquire whether current ASR models support fine-tuning to improve recognition accuracy for an existing low-resource language, specifically Dutch.
If such fine-tuning is feasible, could you kindly provide a general list of the datasets that have already been used for training? This would help us avoid potential data duplication during subsequent training for Dutch and mitigate issues such as overfitting or catastrophic forgetting.
Thanks！