
Exploring Music Collections on Mobile Devices Michael Kuhn Olga Goussevskaia Roger Wattenhofer MobileHCI 2008 Amsterdam, NL
organization by album History • Storage media • Vinyl records • Compact cassetts • Compact discs • An Album is stored on a single physical storage medium • Sequence of songs given by album • Album is typically listened to as a whole Michael Kuhn, ETH Zurich @ MobileHCI 2008
Music today • Huge offer, easily available • filesharing, iTunes, amazon, etc. • Large collections • The entire collection is stored on a single electronic storage medium • Organization by albums (and other lists) is no longer appropriate organize by similarity Michael Kuhn, ETH Zurich @ MobileHCI 2008
Contributions • Vision • Plays songs the user likes • Overview of a collection • Directly on mp3-player (or phone) • Problems on mobile devices • Limited input • Limited output • Limited processing power • Limited memory • Contribution • Use song coordinates that reflect similarity • Proof-of-concept implementation on Android Michael Kuhn, ETH Zurich @ MobileHCI 2008
Music Explorer • www.musicexplorer.org • Webservice that provides 10D coordinates for songs • Similar songs are close to each other in Euclidean space • Similarity information based on co-occurrence data • Currently about 400K songs available • Similarity derived by means ofco-occurrence analysis Michael Kuhn, ETH Zurich @ MobileHCI 2008
Music in Euclidean Space • Performance • Similarity computation comes almost for free: O(1) time • Memory footprint is extremly low: O(1) per song • All information can be saved in the file, no server connection required. • Applications • Trajectories (playlists, ...) • Volumes (region of interest, ...) • etc. coordinates are well suited for mobile applications coordinates are well suited for similarity based organization Michael Kuhn, ETH Zurich @ MobileHCI 2008
Playlist generation • Interpolation between start and end-point • Smooth transition from one style to the other • In reality: 10 dimensions Michael Kuhn, ETH Zurich @ MobileHCI 2008
Similarity-based Navigation • Basic idea: Browse through neighborhood lists • Challenges • Reachability: Entire collection should be reachable from any given starting point • Searchability: It should be possible to reach new regions within few steps Michael Kuhn, ETH Zurich @ MobileHCI 2008
Similarity-based Navigation (Small-World) • J. Kleinberg: The Small-World Phenomenon: An Algorithmic Perspective, STOC’00 • Augmenting a (hyper-)grid with edges following a particular length distribution (d-r, r = #dim) leads to polylog diameter (=>reachability) • Short paths do not only exist, but can be found using local knowledge only (=>searchability) Michael Kuhn, ETH Zurich @ MobileHCI 2008
Similarity-based Navigation (Clustering) • Idea: Cluster similar songs and list clusters instead of single songs • Cover entire collection (=>reachability) • Small clusters for close-by songs • Large clusters for distant regions (=>searchability) Michael Kuhn, ETH Zurich @ MobileHCI 2008
Conclusions and Future Work • Embedding songs into Euclidean space opens many possibilities for mobile applications • We have presented a proof-of-concept Android application that • can create smooth playlists • allows to browse collections based on smilarity • does not require (expensive) connection to a server or DB • Future directions • Visually browsing collections (problem: 10D => 2D) • Playlist generation on the fly • Collaborative features • ... Michael Kuhn, ETH Zurich @ MobileHCI 2008
Thanks for your Attention • Questions? Michael Kuhn, ETH Zurich @ MobileHCI 2008