Mitigating Subgroup Disparities in Multi-Label Speech Emotion Recognition: A Pseudo-Labeling and Unsupervised Learning Approach
The episode discusses a paper on mitigating subgroup disparities in multi-label speech emotion recognition using pseudo-labeling and unsupervised learning. The researchers propose inferring demographic groups from audio without needing explicit labels, which improves fairness across gender, race, and age while maintaining high accuracy.