Meta adversarial learning improves low-resource speech recognition

Yaqi Chen, Xukui Yang, Hao Zhang, Wenlin Zhang, Dan Qu, Cong Chen

Abstract: Highlights•Multilingual meta-learning ignore the imbalance problem across source languages.•Meta adversarial learning can build a more compact semantic space and improves the generalization capability of the model.•Optimized adversarial learning will be more stable by using Wasserstein distance and temporal normalization.•Experiments on target languages verify the effectiveness of meta adversarial learning, especially in very low-resource setting.•Experiments analysis shows the principle and superiority of meta adversarial learning.