Spotify Playlist Demographics
This experiment started with a Spotify playlist and followed its artists into biographical sources, using extracted metadata to build a view of the collection.
Independent development. Python, Spotify, Wikipedia and language-model extraction.
I worked from the artists in a playlist, with Spotify providing the names and Wikipedia providing biographical text that could be turned into structured fields for the analysis, with the cached artist records making the extraction step reusable when another playlist contained the same names.
The pipeline cached extracted information and used GPT to format uncached biographies, then passed those records into plotting code so the playlist could be examined as a group of artists.
The extraction stage makes this exploratory, as biographies can be incomplete and a model can misread them, so the charts need to stay attached to the sources and missing information behind each field.
The catalogue date follows the first preserved commit on 23 March 2024.
Outcome
A preserved retrieval, extraction and plotting workflow. Its demographic fields are source-derived estimates, with no claim that every artist was described completely or correctly.