1. 1 Information Filtering & Recommender Systems (Lecture for CS410 Text Info Systems) ChengXiang Zhai Department of Computer Science University of Illinois, Urbana-Champaign

2. 2 Short vs. Long Term Info Need Short-term information need (Ad hoc retrieval) –“Temporary need”, e.g., info about fixing a bug –Information source is relatively static –User “pulls” information –Application example: library search, Web search,… Long-term information need (Filtering) –“Stable need”, e.g., your hobby –Information source is dynamic –System “pushes” information to user –Applications: news filter, movie recommender, …

3. 3 Part I: Content-Based Filtering (Adaptive Information Filtering)

4. 4 Adaptive Information Filtering Stable & long term interest, dynamic info source System must make a delivery decision immediately as a document “arrives” Filtering System … my interest:

5. 5 A Typical AIF System... Binary Classifier User Interest Profile User Doc Source Accepted Docs Initialization Learning Feedback Accumulated Docs utility func User profile text Linear Utility = 3* #good - 2 *#bad #good (#bad): number of good (bad) documents delivered to user Are the coefficients (3, -2) reasonable? What about (10, -1) or (1, -10)?

6. 6 Three Basic Problems in AIF Making filtering decision (Binary classifier) –Doc text, profile text  yes/no Initialization –Initialize the filter based on only the profile text or very few examples Learning from –Limited relevance judgments (only on “yes” docs) –Accumulated documents All trying to maximize the utility

7. 7 Extend a Retrieval System for Information Filtering “Reuse” retrieval techniques to score documents Use a score threshold for filtering decision Learn to improve scoring with traditional feedback New approaches to threshold setting and learning

8. 8 A General Vector-Space Approach doc vector profile vector Scoring Thresholding yes no Feedback Information Vector Learning Threshold Learning threshold Utility Evaluation

9. 9 Difficulties in Threshold Learning 36.5 Rel 33.4 NonRel 32.1 Rel 29.9 ? 27.3 ? …...  =30.0 Censored data (judgments only available on delivered documents) Little/none labeled data Exploration vs. Exploitation No judgments are available for these documents

10. 10 Empirical Utility Optimization Basic idea –Compute the utility on the training data for each candidate score threshold –Choose the threshold that gives the maximum utility on the training data set Difficulty: Biased training sample! –We can only get an upper bound for the true optimal threshold –Could a discarded item be possibly interesting to the user? Solution: –Heuristic adjustment (lowering) of threshold

11. 11 Beta-Gamma Threshold Learning Cutoff position (descending order of doc scores) Utility 0 1 2 3 … K...  , N, N ,   [0,1] The more examples, the less exploration (closer to  optimal )   Encourage exploration up to  zero

12. 12 Beta-Gamma Threshold Learning (cont.) Pros –Explicitly addresses exploration-exploitation tradeoff (“Safe” exploration) –Arbitrary utility (with appropriate lower bound) –Empirically effective Cons –Purely heuristic –Zero utility lower bound often too conservative

13. 13 Part II: Collaborative Filtering

14. 14 What is Collaborative Filtering (CF)? Making filtering decisions for an individual user based on the judgments of other users Inferring individual’s interest/preferences from that of other similar users General idea –Given a user u, find similar users {u 1, …, u m } –Predict u’s preferences based on the preferences of u 1, …, u m

15. 15 CF: Assumptions Users with a common interest will have similar preferences Users with similar preferences probably share the same interest Examples –“interest is IR” => “favor SIGIR papers” –“favor SIGIR papers” => “interest is IR” Sufficiently large number of user preferences are available

16. 16 CF: Intuitions User similarity (Kevin Chang vs. Jiawei Han) –If Kevin liked the paper, Jiawei will like the paper –? If Kevin liked the movie, Jiawei will like the movie –Suppose Kevin and Jiawei viewed similar movies in the past six months … Item similarity –Since 90% of those who liked Star Wars also liked Independence Day, and, you liked Star Wars –You may also like Independence Day The content of items “didn’t matter”!

17. 17 The Collaboration Filtering Problem u 1 u 2 … u i... u m Users: U Objects: O o 1 o 2 … o j … o n 3 1.5 …. … 2 2 1 3 X ij =f(u i,o j )=? ? The task Unknown function f: U x O  R Assume known f values for some (u,o)’s Predict f values for other (u,o)’s Essentially function approximation, like other learning problems Ratings

18. 18 Memory-based Approaches General ideas: –X ij : rating of object o j by user u i –n i : average rating of all objects by user u i –Normalized ratings: V ij = X ij – n i –Memory-based prediction of rating of object o j by user u a Specific approaches differ in w(a,i) -- the distance/similarity between user u a and u i

19. 19 User Similarity Measures Pearson correlation coefficient (sum over commonly rated items) Cosine measure Many other possibilities!

20. 20 Improving User Similarity Measures Dealing with missing values: set to default ratings (e.g., average ratings) Inverse User Frequency (IUF): similar to IDF

21. 21 Summary Filtering is “easy” –The user’s expectation is low –Any recommendation is better than none –Making it practically useful Filtering is “hard” –Must make a binary decision, though ranking is also possible –Data sparseness (limited feedback information) –“Cold start” (little information about users at the beginning)

22. 22 What you should know Filtering, retrieval, and browsing are three basic ways of accessing information How to extend a retrieval system to do adaptive filtering (adding threshold learning) What is exploration-exploitation tradeoff Know how memory-based collaborative filtering works