Western Music Industry Collaboration Corpus

The dataset contains band-band music collaboration networks based on Discogs data. It contains 39 geographical regions plus a special “global” region. Each region has a deidcated folder with cumulative yearly edge snapshots. Each edge snapshot is an edgelist with the ids of the two connected nodes, the edge weight and significance. Two node attribute files are provided per geographical region: nodes.tsv contains thusands of node attributes per node (including the music genre played, the labels publishing their records, years of activity and more) in sparse format (one line per node-attribute pair); and nodes_viz.tsv which contains a handful of representative attributes in dense format (one line per node). The special global region also contains five pytorch geometric tensors, with different target y variables for different Graph Neural Network prediction tasks.

More information included in the paper currently under review in the Scientific Data journal, will be linked here upon acceptance.