A Glimpse of Recommender Systems on the Web Bin Tan 4/26/07
Classification Method: content-based, collaborative, hybrid Collaborative filtering: user-to-user, item-to-item User & item representation: id, keywords, category, metadata, context (demographics, time, location, social network) Feedback: explicit, implicit
Amazon.com Item-to-item collaborative filtering Feedback collected from purchases, ratings and page views User profile editable Efficient Amazon.com recommendations: item-to-item collaborative filtering
Amazon.com Highlights Customers with Similar Searches Purchased … What Do Customers Buy After Viewing This Item? Better Together Buy this item with … today! Customers who bought this item also bought … Explore similar items: more like this / by category Customers viewing this page may be interested in these Sponsored Links Rate this item to improve your recommendations Today’s Recommendation For You Category Tags Improve Your Recommendations Update your Amazon history to improve your recommendations
Findory.com (52,002) Personalized News Clickthrough as feedback
LibraryThing.com (8,939) Personal library management Add a book by searching catalogs of 70 online libraries Share book rating, tags, reviews Find people with similar books Get book recommendations
Other Popular Sites Last.fm (music) 350 iLike.com (music) 2,322 RateYourMusic.com 5,463 FilmAffinity.com 7,443 Douban.com (books, movies, music) 1,485
StumbleUpon Firefox / IE plug-in Recommend web pages, photos, videos, news Feedback: user-selected categories, item rating (thumb-up/down) Social network features
References Wikipedia: Collaborative Filtering Toward the Next Generation of Recommender Systems: A Survey of the State-of-the-Art and Possible Extensions