Mining Facebook data for music recommendations

October 11, 2018

A group of friends maintains a Facebook group for music recommendations. Every day, someone comes up with a hashtag and the community replies with links to songs, mostly from YouTube. The hard part of the game was, however, to gather all the songs at the end of the day, since on good days we could end up with more than 100 songs recommended by musicians and experts.

My proposed solution to the problem was to use the FacebookR package for R to mine all the information in the group. One thing that I noticed, though, was all the information that you’re actually able to mine by using the Facebook API about publications and reactions. The information was transformed and organized by hashtag. The second part of the project involved using YouTube’s API to retrieve not-so-clean artist and song information. Finally, I “manually” created playlists for each hashtag and put it online on a static web page on a test server for the group. I’ve successfully used the playlists at parties in a few countries around the world, so I’m happy with the result.

You can see the source code here.