ICLC 2023 Catalogue

Reproducible Musical Analysis of Live Coding Performances Using Information Retrieval: A Case Study on the Algorave 10th Anniversary

Georgios Diapoulis, Martin Carlé

Was presented at:

Live Stream Recording (YouTube)

Publication: https://doi.org/10.5281/zenodo.7843813


We present a reproducible music information retrieval (MIR) study on 133 performances from the 10th anniversary of Algorave. Our aim in this paper is to provide a reproducible framework for computational analysis of musical performances. Here, we present a tool for analysing acoustical characteristics and for visualizing the musical structure from performances of one algorave event. Our musical analysis of the live coding performances highlights the musical diversity within the live coding community to a broader scientific audience. At the same time, we expect that the algoravers will gain insights on their own musical practices through the computational analysis of the musical structure of their performances. In concerning ourselves with reproducibility, our intention is to motivate more researchers to analyse musical practices of other under-represented music communities. As a basic tool for reproducibility we construct a pipeline for analysing performances using Python within a Jupyter notebook. To make this reproducible on different computers we wrapped the whole workflow setup into a docker image. We represent the results of our analysis as a series of plots of different kinds. These plots present both overviews of the entire repertory in compact form, and comparisons of individual pieces in more detail. In learning one can use such visualization as a means for raising awareness on one's evolution of the musical outcome. In performance this visualization can be developed to a real-time and possibly an interactive tool which informs the coder about the musical outcome of a live set on-the-fly. Finally, we reflect on how and to what extent such MIR studies can provide valuable insights in live coding performance practices, while also considering the limitations faced when dealing with such large parameter spaces in human machine musicianship.