1 / 23

The Benefit of Using Tag-Based Profiles

The Benefit of Using Tag-Based Profiles. Claudiu Firan, Wolfgang Nejdl, Raluca Paiu 5 th Latin American Web Congress, 2007. Music Recommendation. Personal Music. Community Data. Challenges. New Approach. Personal Music. Personal Tags. Community Data. Why Use Tags?. Tags are:

naava
Download Presentation

The Benefit of Using Tag-Based Profiles

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The Benefit of Using Tag-Based Profiles Claudiu Firan, Wolfgang Nejdl, Raluca Paiu 5th Latin American Web Congress, 2007

  2. Music Recommendation PersonalMusic CommunityData

  3. Challenges

  4. New Approach PersonalMusic PersonalTags CommunityData

  5. Why Use Tags? • Tags are: • Written chaotically • Not verified • Unstructured • Heterogeneous • Unreliable • But if many, the correct ones arise • “Wisdom of the masses”

  6. Last.fm – “The Social Music Revolution” Track Similar Tracks Track Usage Info Tags(with weight) Artist Similar Artists User Comments Albums

  7. Tracks, Tags, and Profiles

  8. User Profiles • weight=preference(user,item)

  9. Track-based Profiles (TR) <tracki, weighti> … • preference(user,track) = • log(user_track_#listened) TR

  10. Track-Tag-based Profiles (TT) <tagi, weighti> … • preference(user,tag) = • log( Σi( • log(user_tracki_#listened) ∙ • log(user_tag_tracki_#tagged))) • [∙ ITF(tag)] • ITF = Inverse Tag Frequency • With: TTI • Without: TTN TTI TTN

  11. Tag-based Profiles (TG) <tagi, weighti> … • preference(user,tag) = • log(user_tag_#used) TG

  12. User Profiles from Personal MP3s • Read personal playlist from PC • Match MP3s against our database • Add overall average usage information values

  13. Collaborative Filtering vs. Search

  14. Track- & Tag-based Recommendations Collaborative Filtering <tracki, weighti> … <tagi, weighti> …

  15. Tag-based Search <tagi, weighti> …

  16. Algorithms

  17. Experiments & Outcome

  18. Last.fm Crawled Data • 317,058 tracks • 21,177 tags (most prominent ones are music genres) • 289,654 users 12,193 listened at least 50 tracks and used at least 10 tags

  19. Experimental Setup • Create user profiles • 18 subjects • 658 tracks on average in user profile (not statistically significant in influencing algorithm outcome) • Run algorithms • 7 algorithms • 10 recommended items per algorithm per user • Two scores • Quality of recommendation [0-2]  NDCG • Novelty of recommendation [0-2]  Average

  20. Results • STG: • Lower popularity • Higher quality CFTR: Baseline • NDCG – Novelty: • High inverse correlation • Pearson c = -0.987 • STTI & STTN: • Huge improvement • Statistically significant

  21. Gain over the Baseline (CF on Tracks)

  22. Conclusions • CF on tag-based profiles worse than CF on track-based profiles • Search with tags improved recommendation performance substantially • 44% increase in quality • Instant results – virtually no time delay • No cold start problem • Tag-based profiles work also with less rich music repositories • Results probably influenced by the consistent tag usage on Last.fm: mostly genres

  23. Thank You!

More Related