“Spot” these PageRank Values
The popular music streaming service, Spotify has a relatively new feature called Discover Weekly that many users have come to love, so much in fact that social media blew up in panic when a bug delayed the release of the playlists a few weeks ago. Discover Weekly is a personalized playlist that is made specifically for every user: updated every week, the playlist of two hours of specifically tailored music is released every Monday morning. Now what makes Discover Weekly different, or rather, better, than other major streaming services?
Unlike Apple Music and many other services, Spotify doesn’t rely on human curators to conjure up static playlists for their users. In his article, Ben Popper describes how Spotify’s major competitors stacked up in terms of their personalized recommendations for him. Not surprisingly, he found that none of them truly compares to Discover Weekly, as most of them use basic collaborative filtering and impersonal fixed playlists. What sets Discover Weekly apart from the rest is its revolutionary use of rather sophisticated algorithms. Popper documented his visit to Spotify’s New York office to find out just how these algorithms work. He found that Spotify builds a personalized taste profile for you by assigning scores to certain artists based on the frequency at which they’re played and on how much you seem to like them. By also bringing in your most frequently played genres, the algorithms find other users whose playlists and frequently played music match up to yours, and pick out songs and artists that these other users have on their playlists that you haven’t listened to yet. Those songs are then filtered through using your taste profile so that its almost guaranteed that you will enjoy them, and these songs finally released as your customized Discover Weekly playlist.
This system embodies the PageRank algorithm that we’ve learned in class. As we know, PageRank essentially works by treating links to certain websites as votes and claiming the sites with the most “votes” as the best sources on certain topics. Google utilizes PageRank to determine which sites come up first in a specific search: the more votes a site has, the more likely Google will refer you to it. Spotify’s algorithm for creating its Discover Weekly playlists work in the same way, as every time a user with a similar taste profile to yours plays a potential song for your Discover Weekly playlist, it gets a vote. The more votes a song gets, the more likely you’ll like it, thus the more likely it will make it onto your playlist. This system seems to work best because although using an algorithm seems to be too impersonal for something as emotional as music, Spotify’s specific algorithm relies on other people to essentially do the song selection. However unlike other music streaming services that use outside employers, Spotify’s method relies on other actual users that share the same interests as you, assuring success of the algorithm and satisfaction of the users.
Article referenced: http://www.theverge.com/2015/9/30/9416579/spotify-discover-weekly-online-music-curation-interview