http://www.last.fm

Last.fm is considered to be the first of its kind, and has been the most successful in its area of service. Essentially, it is a networking service based upon music. Last.fm essentially learns the users preferences in music in a variety of ways, and then creates a radio station based off of the music the user listens to. The most common way that it gets this information is through what it calls “scrobbling”—whenever a user listens to music using their preferred media player, it will take that information and add it to the users profile, where this information will later be used to determine what will be played on the users radio station.

Once the information has been gathered, the networking process comes into play. We can consider users and artists to be nodes on a graph of the network, with edges connecting users that enjoy the same artists, and nodes between artists if people who enjoy the one artist’s music also enjoy another artist’s music. This is how the basic user’s radio station is formed—once they begin listening to music, Last.fm finds music by similar artists—those that are considered the most similar (to an artist) are those which have to most amount of common listeners (to the same artist).

For the most part, this system is effective in helping users find new music, since people who usually listen to a certain song/artist will typically listen to other music that is similar. However, with the most popular artists, there is obviously some problem with the system. Consider an artist like U2. With such a huge fanbase, it seems unlikely that users will have similar preferences in music, or if they do, they will merely include other “Top 40” artists, with no true similarity to the artist in question, U2. This can be seen from their similar artists page on Last.fm, where we find that the top 3 most common artists are Coldplay, R.E.M., and The Police. While these artists do have some things in common, they are not truly the most similar artists to U2 when we consider only their music.

Last.fm has a variety of other options that allow users to access their network in different ways. Each user has a unique “Neighborhood,” i.e., the list of users who listen to similar music. This system is also based upon the same similar artists network, so it can use the same network graph. One can listen to tracks similar to an artist, or similar to another users preferences (such as someone in their “Neighborhood”). However, there are other options that are based upon a different networking system. One can use a “tag” based system that will play songs by artists who have similar “tags.” These “tags” are most commonly used to put the artist’s genre. However, they are also used to put the mood of the artists music, the instrumentation used, their popularity, or even something as arbitrary as “water” (users who put this refer to artists whose names are related to water or who have popular songs about some kind of water). This network has nodes for tags and artists, and an artist who has been given a tag has an edge between themselves and the tag.