Closing the Valence Gap in Emotion Recognition
2021 has been an exciting year for our researches working on the recognition of emotions from speech. Benefiting from the recent advances in transformer-based architectures, we have for the first time built models that predict valence with a similar high precision as arousal.