This project will demonstrate the application of data science and AI techniques to music analysis, for application in digital musicology, creative industries and cultural heritage. Specific focus is on addressing music analysis, multimodal analysis of performance data and music in the archive.
Explaining the science
'Music Information Retrieval' has been a thriving domain for over a decade, and exemplifies many aspects of best practice - for example in analytics, community engagement, approaches to reproducibility and data sharing, and infrastructure (e.g. running code remotely over datasets that cannot be shared directly). As such it provides well defined challenges in data science and in working with archives it's ready for the application of emerging data science and AI techniques, and it also promotes useful practices for data scientists in other domains.
- Development of analytics over music content, in order to support musicological analysis and composition of new works.
- Multimodal analysis of performance reception data, to better understand perception of musical features. This will, for example, further our understanding of the relationship between mathematics and music.
- Demonstration of building links across music content in multiple archives, to help with search, discovery and historical analysis.
The music industry is now digital almost end-to-end, and new recordings are generated all the time. Feature extraction can help throughout the life cycle of music from composition, recording and production to distribution, consumption and re-use. Hence this work can be applied by both music producers and consumers.
There are also creative applications, for example using AI, algorithmic composition and sonification, as well as algorithmic enhancement of instruments.