Zixun (Nicolas) Guo PhD StudentEmail: zixun.guo@qmul.ac.uk Website: https://guozixunnicolas.github.io ProfileResearchProfileProject title: Towards Tonality-Aware Music Understanding: Modeling Complex Tonal Harmony Abstract: Artificial Intelligence (AI) has recently advanced many fields including Music Information Retrieval (MIR), from both the analytic and generative perspective. Besides its powerfulness, bottlenecks exist for AI models. For example, the "blackbox" effect makes the music modelling process unexplainable. It must be admitted that MIR did not start from AI and several pioneer algorithms (e.g., tonality modelling , pattern discovery were developed mainly from music intuition and careful feature engineering. Yet, well-established music theories and music intuitions are often discarded in recent AI-based music models. One could argue that AI should be able to 'learn' the domain knowledge via training but this is often not the case. C4DM theme affiliation: Music Informatics, Machine ListeningResearchResearch Interests:Music Understanding, Automatic Music Transcription, Automatic Music Generation.