1 / 33

Tagommenders : Connecting Users to Items Through Tags

Tagommenders : Connecting Users to Items Through Tags. Shilad Sen Macalester College Jesse Vig , John Riedl GroupLens Research. Tagommenders Analyze user interactions to infer liking (preferences) for tag concepts. Recommend items related to tag concepts liked by users.

nascha
Download Presentation

Tagommenders : Connecting Users to Items Through Tags

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. Tagommenders:Connecting Users to Items Through Tags Shilad Sen Macalester College Jesse Vig, John Riedl GroupLens Research

  2. Tagommenders Analyze user interactions to infer liking (preferences) for tag concepts. Recommend items related to tag concepts liked by users.

  3. Tagommender Goals • Recommend items using just tags. (Delicious) • Improve item recommendations with ratings by by using tags. (LibraryThing / Amazon) • accuracy • flexibility • explainability(Vig, IUI 2009).

  4. Tagommender Flow Chart WALL-E animation robots pixar tag preference inference tag-based recommendation

  5. MovieLens Tagging • Tagging introduced in 2006 • 15,000 distinct tags • 127,000 tag applications: • <user, tag, movie> • 4000 users applied >= 1 tag • 7700 movies with >= 1 tag app

  6. Outline • Tag preference inference • Item recommendation • Auto-tagging and wrap-up

  7. Outline • Tag preference inference • Item recommendation • Auto-tagging and wrap-up

  8. Step 1: Tag Preference Inference ? animation robots pixar • Infer a user’s interest in tags from: • tags user applied • tags user searched for • user’s clicks on movie hyperlinks • user’s movie ratings

  9. 118,017 ratings by 995 users

  10. Preferences for Tags Searched / Applied

  11. Movie-rating algorithm cars

  12. Movie-Rating Algorithm cars 4 of 12 0.8 9 of 38 0.9 1 of 36 0.1

  13. Generative Model: Expressive probabilistic processes. Model movie ratings. Separate model for every user, tag. Bayes-Rating Algorithm

  14. Bayes-Rating Algorithm Jill’s Ratings for animated Movies N(μ=3.8,σ=0.7)

  15. Bayes-Rating Algorithm WALL-E p(t| WALL-E) 1.0 - p(t| WALL-E) t = animation not t all possible normal dists for ratings for animated movies N(μu,σu) N(μu,t,σu,t) N(μ=2.0,σ=1.0) N(μ=4.0,σ=0.5)

  16. Bayes-Rating Algorithm All movies m rated by Jill tagged with animation t = animation not t all possible normal dists for ratings for animated movies Toy Story WALL-E Shrek

  17. Outline • Tag preference inference • Item recommendation • Auto-tagging and wrap-up

  18. Tagommender Flow Chart WALL-E animation robots pixar tag preference inference tag-based recommendation

  19. Standard machine learning problem With / without ratings Sixstandard recommender baselines Evaluate predictive performance Step #2: Tag-Based Recommendation

  20. Outline • Tag preference inference • Item recommendation • Auto-tagging and wrap-up

  21. Using Tag Preferences for Tag Inference

  22. Top 10 Inferred Tags Not Already Applied

  23. Tag preference inference: Systems can infer user preferences for tags. Item ratings help tag prefinference. Tag prefs can be used for auto-tagging. Tagommenders outperform traditional recommenders: Without ratings: moderate edge (10%). With ratings: slight edge (2%). Summary of Tagommenders

  24. Future Work Alternative modalities for tags. Quality vs. preference. Thank You! GroupLens. MovieLens users. NSF grants IS 03-24851 and IIS 05-34420. Macalester College.

  25. Shilad Sen ssen@macalester.edu (photo by flickr user SantiMB)

More Related