Less can be More: Towards a Parameter-Efficient Fine-Tuning of Wav2Vec2 XLSR for Low-Resource Cape Verdean Creole ASR

Mateus Neves Andrade, Mouhamadou Lamine Ba, Idy Diop, Arlindo Oliveira da Veiga


Abstract
Automatic Speech Recognition (ASR) for low-resource languages remains challenging due to limited annotated data and high linguistic variability. In this work, we investigate parameter-efficient fine-tuning strategies for Cape Verdean Creole ASR using the Wav2Vec 2.0 XLSR model. We evaluate the impact of structured layer freezing on model performance, training stability, and computational efficiency. Experiments conducted on a newly curated Santiago-dialect dataset show that full fine-tuning achieves the best absolute performance (WER 0.212, CER 0.120). However, several freezing configurations achieve comparable recognition performance while substantially reducing the number of trainable parameters and exhibiting more stable convergence. These results highlight a trade-off between adaptability and efficiency, showing that selective freezing can serve as an effective regularization strategy in low-resource settings. This work provides practical insights into parameter-efficient adaptation for under-resourced Creole languages.
Anthology ID:
2026.rail-1.7
Volume:
Proceedings of Resources for African Indigenous Languages (RAIL) 2026 @ LREC 2026
Month:
May
Year:
2026
Address:
Palma, Mallorca (Spain)
Editors:
Muzi Matfunjwa, Mmasibidi Setaka, Rooweither Mabuya, Menno van Zaanen
Venues:
RAIL | WS
SIG:
Publisher:
ELRA Language Resources Association (ELRA)
Note:
Pages:
62–71
Language:
External URL:
https://lrec.elra.info/lrec2026-ws-rail-07
DOI:
10.63317/4b6ij7awds27
Bibkey:
Cite (ACL):
Mateus Neves Andrade, Mouhamadou Lamine Ba, Idy Diop, and Arlindo Oliveira da Veiga. 2026. Less can be More: Towards a Parameter-Efficient Fine-Tuning of Wav2Vec2 XLSR for Low-Resource Cape Verdean Creole ASR. In Proceedings of Resources for African Indigenous Languages (RAIL) 2026 @ LREC 2026, pages 62–71, Palma, Mallorca (Spain). ELRA Language Resources Association (ELRA).
Cite (Informal):
Less can be More: Towards a Parameter-Efficient Fine-Tuning of Wav2Vec2 XLSR for Low-Resource Cape Verdean Creole ASR (Andrade et al., RAIL 2026)
Copy Citation: