As the amount, availability, and use of online music increase, music recommendation becomes an important field of research. Collaborative, content-based and case-based recommendation systems and their hybrids have been used for music recommendation. There are already a number of online music recommendation systems. Although specific user information, such as, demographic data, education, and origin have been shown to affect music preferences, they are usually not collected by the online music recommendation systems, because users would not like to disclose their personal data. Therefore, user models mostly contain information about which music pieces a user liked and which ones s/he did not and when.
TopBackground
Widespread use of mp3 players and cell-phones and availability of music on these devices according to user demands, increased the need for more accurate Music Information Retrieval (MIR) Systems. Music recommendation is one of the subtasks of MIR Systems and it involves finding music that suits a personal taste (Typke et.al., 2005). The content search in MIR systems could also be used to identify the music played, for example query-by-humming (Ghias et.al., 1995), to identify suspicious sounds recorded by surveillance equipment, to make content-based video retrieval more accurate by means of incorporating music content, to help theaters and film makers find appropriate sound effects (Typke et.al. 2005), to produce audio notification to individuals or groups (Jung & Heckmann 2006).
Music recommendation tasks could be in the form of recommending a single album/song (Logan 2004) or a series of them as in playlist generation (Aucouturier & Pachet, 2002; Alghoniemy & Tewfik, 2000). In addition to containing interesting songs for the user or the user group, a playlist have to obey certain conditions, such as containing all different songs, having a certain duration, having continuity and progression from one song to the next (Aucouturier & Pachet, 2002). Therefore, playlist generation is a harder task than single music item recommendation.
The songs to recommend could contain the audio or MIDI content, as well as, genre, artist, lyrics and other information. The audience of a music recommendation system could be a single person or a group of people (Baccigalupo & Plaza, 2007; McCarthy et.al. 2006). The audience or the songs could be dynamic or mostly static. Depending on these task and user requirements, different algorithms have to be employed for music recommendation. Yahoo Launch!, Last.FM, Pandora (music genome project), CDNow, Audioscrobbler, iRate, MusicStrands, inDiscover (Celma et.al., 2005) are some of the music recommendation projects.