Optimize Wav2vec2s Architecture for Small Training Set Through Analyzing its Pre-Trained Models Attention Pattern.

Liu Chen, Meysam Asgari, Hiroko H. Dodge

Browse the full ICASSP paper archive.