Improving Spoken Language Identification with Map-Mix

June 2023 Shangeth Rajaa, Kriti Anandan, Swaraj Dalmia, Tarun Gupta, Eng Siong Chng ICASSP 2023 — IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 1–5

#Voice AI #Speech #Language Identification #Data Augmentation

Abstract

We address dialect classification in low-resource settings using a pre-trained multilingual XLSR model. We introduce Map-Mix, a data augmentation technique that uses model training dynamics to improve sampling for latent mixup operations. The method achieves approximately 2% improvement in weighted F1 scores compared to random mixup and produces better-calibrated models.