Genre Co-Occurrence in the Free Music Archive

A heat-map of the of the frequency with which the top 100 genres in the FMA show up together in the work of the same artist.

This heatmap visualizes relationships between genres of songs included in the Free Music Archive (FMA).

Relationships were tracked by analyzing the ID3 metadata of all of the songs in the archive, and tracking how frequently songs of one genre occurred alongside songs of another genre in the body of work by the same artist. In this way, we could tally, for instance, how many songs categorized as "jazz" were produced by artists who also wrote songs categorized as "electronic."

Multiple co-occurrences between the same genres within the same artist's oeuvre were tallied individually to reflect a stronger connection in, for instance, an artist's body of work that featured many pieces focused within two particular genres. For example, if an artist's ouevre contained 3 songs labelled "jazz" and two labeled "electronic," each jazz song co-occurs with two electronic songs, resulting in a co-occurence value, for that artist, and those two genres, of 6 (effectively the product of both genres' occurrence in a single artist's body of work). By repeating this process for all artists in the Free Music Archive, and summing the values, we get a total co-occurrence value for this pair of genres. And by completing this process for every possible genre pairing in the 1,203 genre categories present in the FMA, we can generate a map of genre co-occurrences.

The raw tallies resulted in more common genres also appearing to have stronger affinities. To adjust for this, all co-occurence tallies were divided by the average incidence in the FMA dataset of the two genres being compared. For example, the adjusted co-occurence value of "pop" and "electronic" is the total co-occurence value of these two genres, calculated as described above, divided by the average of the total number of pop and electronic songs in the FMA.

This heatmap only tracks the top 100 most common genres in the FMA. The FMA includes metatags for 1,203 distinct genre categories. The full heatmap data, including all genres in the FMA, can be downloaded here:

genre_coincidence_adjusted-20201208-204319.xlsx (Excel, color-coded)
genre_coincidence_adjusted-20201208-204319.csv (CSV)