Original Paper
The purpose of this repo is to further investigate the ideas exposed in the Meta-Prod2Vec paper, in particular on the topic of homogenous representation learning of hetergenous data. With no particular use case in mind, any idea proposed should be tried as to compare with other approaches.

Note there is no reference implementation of MetaProd2Vec, and as so, there's no clear baseline to compare with.

MSD @ github
Unverified Spotify Playlists
Julián Bayardo's talk @ PyData 2018
Collab Notebook allegedly used in such talk
CS Cornell Playlist Dataset 8Track Playlist Dataset 30MusicDataset idomaar data format

Run the setup.sh script provided to create the missing repo structure and download the necessary data.

So, it'd be cool to have spotify data for research (playlists, sessions). So, since no public dataset was found, maybe we could scrap the data? Yeah, well, that ain't happenning. Check the developer ToS, section IV, article 1. General restrictions, point g.