Adaptive User Modeling for Content-Based Music Retrieval

  • Kay Wolter
  • Christoph Bastuck
  • Daniel Gärtner
Part of the Lecture Notes in Computer Science book series (LNCS, volume 5811)

Abstract

An approach to adapt a content-based music retrieval system (CBMR system) to the user is presented and evaluated. Accepted and rejected songs are gathered to extract the user’s preferences. To compare acoustic characteristics of music files, profiles are introduced. These are based on result lists. Each result list is created by a classifier and sorted accordingly to the similarity of the given seed song. To detect important characteristics, the accepted and rejected songs are clustered with k-means. A score for each candidate song is specified by the distance to the mean values of the obtained clusters. The songs are proposed by creating a playlist, which is sorted by the score. Songs accepted by the listener are used to query the CBMR system for new songs and thus extract additional profiles. It is shown that incorporating relevance feedback can significantly improve the quality of music recommendation. The L2 distance is suitable to determine similarities between profiles of regarded songs. Introducing more than one query song during the recommendation process can further improve the quality.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Mandel, M., Poliner, G., Ellis, D.: Support Vector Machine Active Learning for Music Retrieval. Multimedia Systems 12, 3–13 (2006)CrossRefGoogle Scholar
  2. 2.
    Pampalk, E., Pohle, T., Widmer, G.: Dynamic Playlist Generation Based on Skipping Behaviour. In: International Conference on Music Information Retrieval, vol. 6, pp. 634–637 (2005)Google Scholar
  3. 3.
    Logan, B.: Music Recommendation From Song Sets. In: International Conference on Music Information Retrieval, vol. 5, pp. 425–428 (2004)Google Scholar
  4. 4.
    Lampropoulos, A., Sotiropoulos, D., Tsihrintzis, G.: Individualization of Music Similarity Preception via Feature Subset Selection. In: IEEE International Conference on Systems, Man & Cybernetics, vol. 1, pp. 552–556 (2004)Google Scholar
  5. 5.
    Peeters, G.: A Large Set of Audio Features for Sound Description (Similarity and Classification) in the CUIDADO Project. Technical Report, IRCAM, Paris, France (2004)Google Scholar
  6. 6.
    Dittmar, C., Bastuck, C., Gruhne, M.: Novel Mid-Level Audio Features for Music Similarity. In: International Conference on Music Communication Science, pp. 38–41 (2007)Google Scholar
  7. 7.
    Bastuck, C.: Weiterentwicklung eines Verfahrens zur automatischen Bestimmung musikalischer Ähnlichkeit. Master’s Thesis, University Siegen (2006)Google Scholar
  8. 8.
    Dwork, D., Ravi, S., Naor, M., Sivakumar, D.: Rank Aggregation Methods for the Web. In: Proceedings of World Wide Web, vol. 10, pp. 613–622 (2001)Google Scholar
  9. 9.
    Kanungo, T., Mount, D., Netanyahu, N., Piatko, C., Silverman, R., Wu, A.: An Efficient k-Means Clustering Algorithm: Analysis and Implementation. IEEE Transactions on Pattern Analysis and Machine Intelligence 24, 881–892 (2002)CrossRefGoogle Scholar
  10. 10.
    Lloyd, S.: Least Squares Quantization in PCM. IEEE Transactions in Information Theory 28, 129–137 (1982)MATHCrossRefMathSciNetGoogle Scholar
  11. 11.
    Geleijnse, G., Schedl, M., Knees, P.: The Quest for Ground Truth in Musical Artist Tagging in the Social Web Era. In: International Conference on Music Information Retrieval, vol. 8, pp. 525–530 (2007)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  • Kay Wolter
    • 1
  • Christoph Bastuck
    • 1
  • Daniel Gärtner
    • 1
  1. 1.Fraunhofer Institute for Digital Media TechnologyIlmenauGermany

Personalised recommendations