Maurice Grant '13, Adeesha Ekanayake '14, and Prof. Doug Turnbull (Computer Science) recently had their peer-reviewed paper "MeUse: Recommending Internet Radio Stations" accepted for publication in the proceedings of the 14th International Conference on Music Information Retrieval. The conference will be held in Curitiba, Brazil from November 4-8.
Prof. Turnbull is also a co-author on a second paper entitled "Taste Over Time: The Temporal Dynamics of User Preferences" which will also be presented at the conference.
Both papers can be found at:
A demo of MeUse can be found at:
MeUse: Recommending Internet Radio Stations
Maurice Grant, Adeesha Ekanayake, Douglas Turnbull
International Symposium on Music Information Retrieval (ISMIR '13)
Curitiba, Brazil, November 2013
In this paper, we describe a novel Internet radio recommendation system called MeUse. We use the Shout- cast API to collect historical data about the artists that are played on a large set of Internet radio stations. This data is used to populate an artist-station index that is similar to the term-document matrix of a traditional text-based in- formation retrieval system. When a user wants to find stations for a given seed artist, we check the index to deter- mine a set of stations that are either currently playing or have recently played that artist. These stations are grouped into three clusters and one representative station is selected from each cluster. This promotes diversity among the stations that are returned to the user. In addition, we provide additional information such as relevant tags (e.g., genres, emotions) and similar artists to give the user more contextual information about the recommended stations. Finally, we describe a web-based user interface that provides an interactive experience that is more like a personalized Internet radio player (e.g., Pandora) and less like a search engine for Internet radio stations (e.g., Shoutcast). A small- scale user study suggests that the majority of users enjoyed using MeUse but that providing additional contextual in- formation may be needed to help with recommendation transparency.