How Far Can Chord-Symbol Time-Series Adaptation Carry Genre Identity? Capabilities and Boundaries in Multi-Genre Chord-Symbol Modeling
A new research paper explores the effectiveness of adapting a Music Transformer model for various musical genres. The study tested five adaptation methods, including LoRA and IA3, across eleven genres, finding that all methods improved chord prediction accuracy. However, the research concluded that while chord symbols can enhance genre-specific harmonic prediction, they are insufficient on their own to fully capture a genre's identity. AI
IMPACT Demonstrates that AI can learn and adapt to genre-specific harmonic patterns in music, though full genre identity remains elusive with chord symbols alone.