Presentation on theme: "Recommender Systems. Finding Trusted Information How many cows in Texas?"— Presentation transcript:
Finding Trusted Information How many cows in Texas?
Outline What are Recommender Systems? How do they work? How can we integrate social information / trust? What are some applications?
How do they work? Two main methods Find things people similar to me like Find things similar to the things I like
Example with People Who is a better predictor for Alice? Compute the correlation: Bob 0.26 Chuck: 0.82 Recommend a rating of Vertigo for Alice Bob rates it a 3 Chuck rates it a 5 (0.26 * * 5) / ( ) = 4.5
Item similarity Methods are more complex Computed using features of items E.g. genre, year, director, actors, etc. Some sites use a very nuanced set of features
Music DNA arrangement - the selection and adaptation of a composition or parts of a composition to instruments for which it was not originally designed beat - the regular pulse of music form - the structure of a composition, the frame upon which it is constructed; based upon repetition, contrast, and variation harmony - the concordant (or consonant) combination of notes sounded simultaneously to produce chords lyrics - the words of a song melody - a succession of tones comprised of mode, rhythm, and pitches so arranged as to achieve musical shape
More DNA orchestration - the art of arranging a composition for performance by an instrumental ensemble rhythm - the subdivision of a space of time into a defined, repeated pattern syncopation - deliberate upsetting of the meter or pulse of a composition by means of a temporary shifting of the accent to a weak beat or an off-beat tempo - the speed of the rhythm of a composition vamping - to extemporize the accompaniment to a solo voice or instrument voice - the production of sound from the vocal chords, often used in music; falls into six basic categories defined by pitch, ranging, from bottom to top, Bass, Baritone, Tenor, Contralto, Mezzo Soprano, and Soprano
How good is a Recommender System? Generally: error Error = My rating - Recommended Rating Do this for all items and take the average Need alternative ways of evaluating systems Serendipity over accuracy Diversity
Trust in Recommender Systems If we have a social network, can we use it to build trusted recommender systems? Where does the trust come from? How can we compute trust? Some example applications
1. Your Favorite Movie 2. Your Least Favorite Movie 3. Some Mediocre Movie 4. Some Mediocre Movie 5. Some Mediocre Movie 6. Some Mediocre Movie 7. Some Mediocre Movie 8. Some Mediocre Movie 9. Some Mediocre Movie 10. Some Mediocre Movie In-Class Exercise
Sample Profile Knowing this information, how much do you trust User 7 about movies?
Factors Impacting Trust Overall Similarity Similarity on items with extreme ratings Single largest difference Subjects propensity to trust
Advogato Peer certification of users Master : principal author or hard-working co-author of an "important" free software project Journeyer: people who make free software happen Apprentice: someone who has contributed in some way to a free software project, but is still striving to acquire the skills and standing in the community to make more significant contributions Advogato trust metric applied to determine certification
Advogato Website Certifications are used to control permissions Only certified users have permission to comment Combination of certifications and interest ratings of users blogs are used to filter posts
MoleSkiing (note: in Italian) Ski mountaineers provide information about their trips Users express their level of trust in other users The system shows only trusted information to every user Uses MoleTrust algortihm
FilmTrust Movie Recommender Website has a social network where users rate how much they trust their friends about movies Movie recommendations are made using trust Recommended Rating = Weighted average of all ratings, where weight is the trust (direct or inferred) that the user has for the rater
Challenges Can these kinds of approaches create problems? Recommender Systems - recommending items that are too similar Trust-based recommendations - preventing the user from seeing other perspectives
Conclusions Recommender systems create personalized suggestions to users Social trust is another way of personalizing content recommendations Connect social relationships with online content to highlight the most trustworthy information Still many challenges to doing this well