Researchers have developed a novel multitask learning framework to unify music emotion recognition (MER) across different types of emotion labels, including categorical (e.g., happy, sad) and dimensional (e.g., valence-arousal). The framework integrates musical features like key and chords with MERT embeddings and employs knowledge distillation to transfer learning from individual datasets to a generalized student model. Experiments on datasets such as MTG-Jamendo, DEAM, PMEmo, and EmoMusic demonstrated that this approach significantly improves performance, outperforming state-of-the-art models on the MTG-Jamendo dataset. AI
RANK_REASON The cluster contains an academic paper detailing a new framework for music emotion recognition. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
