Diversified Retrieval as Structured Prediction

1. Diversified Retrieval as Structured Prediction Redundancy, Diversity, and Interdependent Document Relevance (IDR ’09) SIGIR 2009 Workshop Yisong Yue Cornell University Joint work with Thorsten Joachims

2. Need for Diversity (in IR) Ambiguous Queries Different information needs using same query “Jaguar” At least one relevant result for each information need Learning Queries User interested in “a specific detail or entire breadth of knowledge available” [Swaminathan et al., 2008] Results with high information diversity

3. Optimizing Diversity • Interest in information retrieval • [Carbonell & Goldstein, 1998; Zhai et al., 2003; Zhang et al., 2005; Chen & Karger, 2006; Zhu et al., 2007; Swaminathan et al., 2008] • Requires inter-document dependencies • Impossible with standard independence assumptions • E.g., probability ranking principle • No consensus on how to measure diversity.

4. This Talk • A method for representing and optimizing information coverage • Discriminative training algorithm • Based on structural SVMs • Appropriate forms of training data • Requires sufficient granularity (subtopic labels) • Empirical evaluation

5. Choose top 3 documents • Individual Relevance: D3 D4 D1 • Pairwise Sim MMR: D3 D1 D2 • Best Solution: D3 D1 D5

6. How to Represent Information? • Discrete feature space to represent information • Decomposed into “nuggets” • For query q and its candidate documents: • All the words (title words, anchor text, etc) • Cluster memberships (topic models / dim reduction) • Taxonomy memberships (ODP) • We will focus on words and title words.

7. Weighted Word Coverage • More distinct words = more information • Weight word importance • Will work automatically w/o human labels • Goal: select K documents which collectively cover as many distinct (weighted) words as possible • Budgeted max coverage problem (Khuller et al., 1997) • Greedy selection yields (1-1/e) bound. • Need to find good weighting function (learning problem).

8. Example Document Word Counts Marginal Benefit

9. Example Document Word Counts Marginal Benefit

10. How to Weight Words? • Not all words created equal • “the” • Conditional on the query • “computer” is normally fairly informative… • …but not for the query “ACM” • Learn weights based on the candidate set • (for a query)

11. Prior Work • Essential Pages [Swaminathan et al., 2008] • Uses fixed function of word benefit • Depends on word frequency in candidate set • - Local version of TF-IDF • - Frequent words low weight • (not important for diversity) • - Rare words low weight • (not representative)

12. Linear Discriminant • x = (x1,x2,…,xn) - candidate documents • v – an individual word • We will use thousands of such features

13. Linear Discriminant • x = (x1,x2,…,xn) - candidate documents • y – subset of x (the prediction) • V(y) – union of words from documents in y. • Discriminant Function: • Benefit of covering word v is thenwT(v,x)

14. Linear Discriminant • Does NOT reward redundancy • Benefit of each word only counted once • Greedy has (1-1/e)-approximation bound • Linear (joint feature space) • Suitable for SVM optimization

15. More Sophisticated Discriminant • Documents “cover” words to different degrees • A document with 5 copies of “Thorsten” might cover it better than another document with only 2 copies.

16. More Sophisticated Discriminant • Documents “cover” words to different degrees • A document with 5 copies of “Thorsten” might cover it better than another document with only 2 copies. • Use multiple word sets, V1(y), V2(y), … , VL(y) • Each Vi(y) contains only words satisfying certain importance criteria. • Requires more sophisticated joint feature map.

17. Conventional SVMs • Input: x (high dimensional point) • Target: y (either +1 or -1) • Prediction: sign(wTx) • Training: subject to: • The sum of slacks upper bounds the accuracy loss

18. Structural SVM Formulation • Input: x (candidate set of documents) • Target: y (subset of x of size K) • Same objective function: • Constraints for each incorrect labeling y’. • Scoreof best y at least as large as incorrect y’ plus loss • Requires new training algorithm [Tsochantaridis et al., 2005]

19. Weighted Subtopic Loss • Example: • x1 covers t1 • x2 covers t1,t2,t3 • x3 covers t1,t3 • Motivation • Higher penalty for not covering popular subtopics • Mitigates effects of label noise in tail subtopics

20. Diversity Training Data TREC 6-8 Interactive Track Queries with explicitly labeled subtopics E.g., “Use of robots in the world today” Nanorobots Space mission robots Underwater robots Manual partitioning of the total information regarding a query

21. Experiments TREC 6-8 Interactive Track Queries Documents labeled into subtopics. 17 queries used, considered only relevant docs decouples relevance problem from diversity problem 45 docs/query, 20 subtopics/query, 300 words/doc Trained using LOO cross validation

22. TREC 6-8 Interactive Track • Retrieving 5 documents

23. Can expect further benefit from having more training data.

24. Moving Forward • Larger datasets • Evaluate relevance & diversity jointly • Different types of training data • Our framework can define loss in different ways • Can we leverage clickthrough data? • Different feature representations • Build on top of topic modeling approaches? • Can we incorporate hierarchical retrieval?

25. References & Code/Data • “Predicting Diverse Subsets Using Structural SVMs” • [Yue & Joachims, ICML 2008] • Source code and dataset available online • http://projects.yisongyue.com/svmdiv/ • Work supported by NSF IIS-0713483, Microsoft Fellowship, and Yahoo! KTC Grant.

26. Extra Slides

27. More Sophisticated Discriminant • Separate i for each importance level i. • Joint feature map  is vector composition of all i • Greedy has (1-1/e)-approximation bound. • Still uses linear feature space.

28. Maximizing Subtopic Coverage • Goal: select K documents which collectively cover as many subtopics as possible. • Perfect selection takes n choose K time. • Basically a set cover problem. • Greedy gives (1-1/e)-approximation bound. • Special case of Max Coverage (Khuller et al, 1997)

29. Learning Set Cover Representations • Given: • Manual partitioning of a space • subtopics • Weighting for how items cover manual partitions • subtopic labels + subtopic loss • Automatic partitioning of the space • Words • Goal: • Weighting for how items cover automatic partitions • The (greedy) optimal covering solutions agree

30. Essential Pages

31. Essential Pages x = (x1,x2,…,xn) - set of candidate documents for a query y – a subset of x of size K (our prediction). Benefit of covering word v with document xi Importance of covering word v • Intuition: • Frequent words cannot encode information diversity. • Infrequent words do not provide significant information [Swaminathan et al., 2008]

32. Structural SVMs

33. Minimizing Hinge Loss Suppose for incorrect y’: Then: [Tsochantaridis et al., 2005]

34. Finding Most Violated Constraint • A constraint is violated when • Finding most violated constraint reduces to

35. Finding Most Violated Constraint • Encode each subtopic as an additional “word” to be covered. • Use greedy prediction to find approximate most violated constraint.

36. Original SVM Problem Exponential constraints Most are dominated by a small set of “important” constraints Structural SVM Approach Repeatedly finds the next most violated constraint… …until set of constraints is a good approximation. Illustrative Example

37. Original SVM Problem Exponential constraints Most are dominated by a small set of “important” constraints Structural SVM Approach Repeatedly finds the next most violated constraint… …until set of constraints is a good approximation. Illustrative Example

38. Original SVM Problem Exponential constraints Most are dominated by a small set of “important” constraints Structural SVM Approach Repeatedly finds the next most violated constraint… …until set of constraints is a good approximation. Illustrative Example

39. Original SVM Problem Exponential constraints Most are dominated by a small set of “important” constraints Structural SVM Approach Repeatedly finds the next most violated constraint… …until set of constraints is a good approximation. Illustrative Example

40. Approximate Constraint Generation Theoretical guarantees no longer hold. Might not find an epsilon-close approximation to the feasible region boundary. Performs well in practice.

41. Approximate constraint generation seems to work perform well.

42. Experiments

43. TREC Experiments • 12/4/1 train/valid/test split • Approx 500 documents in training set • Permuted until all 17 queries were tested once • Set K=5 (some queries have very few documents) • SVM-div – uses term frequency thresholds to define importance levels • SVM-div2 – in addition uses TFIDF thresholds

44. TREC Results

45. Synthetic Dataset • Trec dataset very small • Synthetic dataset so we can vary retrieval size K • 100 queries • 100 docs/query, 25 subtopics/query, 300 words/doc • 15/10/75 train/valid/test split

46. Consistently outperforms Essential Pages