Efficient computation of diverse query results

1. Efficient computation of diverse query results Presenting: Karina Koifman Course : DB Seminar

2. Example

3. Example Yahoo! Autos

4. Maybe a better retrieval

5. Introduction • The article talks about the problem of efficiently computing diverse query results in online shopping applications.

6. The Goal • The goal of diverse query answering is to return a representative set of top-k answers from all the tuples that satisfy the user selection condition

7. The Problem • Users issues query for a product • Only most relevant answers are shown. • Many Duplications

8. Agenda • Existing Solutions • Definition of diversity • Impossibility results of diversity. • Query processing technique.

9. Existing Solutions Existing solutions are inefficient or do not work in all situations. Example: • Obtain all the query results and then pick a diverse subset from these results  doesn’t scale for large data sets.

10. Existing Solutions • Web search engines: first retrieve c × k and then pick a diverse subset from these. • It is more efficient than the previous method. • many duplicates product sale. (inefficient and doesn’t guarantee diversity)

11. Existing Solutions • issuing multiple queries to obtain diverse results:

12. Pro’s\Con’s • The good: • Diversity • The Bad: • Hurts performance • Empty results *There are no Honda Accord convertibles

13. Agenda • Existing Solutions • Definition of diversity • Impossibility results of diversity. • Query processing technique.

14. Diversity Ordering • A diversity ordering of a relation R with attributes A, denoted by , is a total ordering of the attributes in A. • Example: Make ≺ Model ≺ Color ≺ Year ≺ Description ≺ Id

15. The DB example

16. Similarity – SIM(X,Y) Find a result set that minimizes

17. Example - Similarity

18. Prefix

19. Few more definitions • RES(R,Q) of size k • Given relation R and query Q, let maxval =

20. Agenda • Existing Solutions • Definition of diversity • Impossibility results of diversity. • Query processing technique.

21. Impossibility Results • Intuition: IR score of an item depends only on the item and possibly statistics from the entirecorpus, but diversity depends on the other items in the query result set.

22. Inverted Lists Honda cars Honda Car Merged Inverted List:

23. Impossibility Results • Item in an inverted list has a score, which can either be a global score (e.g., PageRank) or a value/keyword -dependent score (e.g., TF-IDF). • The items in each list are usually ordered by their score – so that we could handle top-k queries . • If we assume that we have a scoring function f() that is monotonic- which as a normal assumption for traditional IR system, then the article proofs either it’s not diverse or to inefficient\infeasible.

24. Agenda • Existing Solutions • Definition of diversity • Impossibility results of diversity. • Query processing technique.

25. The DB example

26. The car indexing example

27. One-pass Algorithm Lets say Q looks for descriptions with ‘Low’, with k=3 Honda.Civic.Green.2007.’Low miles’

28. One-pass Algorithm We start from two Civics , then we know that we need only one more so we pick the next Civic

29. One-pass Algorithm Then we look for another in next level (Accord)- no such, because it doesn’t have ‘Low’ in it (also no other in that level).

30. One-pass Algorithm Then we look for another in next level (make)- and prune, This is maximum diverse – we stop here.

31. One-pass Algorithm If we had a Ford, we would continue Ford 0 Focus 0 Black 0 07 0 Low miles

32. Scored One-pass Algorithm Give each car a score , then the query would take this score as parameter- minScore- smallest score in the result set, Choose next next ID by : The smallest ID such that score(id)>=root.minScore. And the algorithm proceeds as before.

33. Probing Algorithm Main idea: to go over all the cars as they were on an axis K=3 K=2 K=1

34. Advantage of bidirectional exploring • “Honda” only has one child,we found it quickly not exploring every option (only civic). • Each time we add a node to the diverse solution we do not have to prune it- unlike the OnePass algorithm.

35. WAND algorithm • WAND is an efficient method of obtaining top-K lists of scored results, without explicitly merging the full inverted lists. • AND(X1,X2,...Xk)≡ WAND(X1,1,X2,1, ...Xk,1,k), • OR(X1,X2,...Xk) ≡ WAND(X1,1,X2,1, ...Xk,1,1). • To obtain k best results the operator uses the upper bounds of maximum contribution, and temp threshold. WAND(X1,UB1,X2,UB2,...,Xk ,UBk, θ)

36. Scored Probing Algorithm We use the WAND algorithm- to obtain the top-k list. Next step is marking all possible nodes to add- as MIDDLE. we also maintain a heap – for a node with minimum child. Each step we move nodes from tentative to useful .

37. Experiments MultQ – rewriting the query as multiple queries and merging their results. Naïve – all the results of a query Basic - just first k answers – without diversity. OnePass , Probe – our algorithms U = unscored S = scored

38. Experiments

39. Experiments

40. Conclusions • Formalized diversity in structured search and proposed inverted-list algorithms. • The experiments showed that the algorithms are scalable and efficient. • In particular, diversity can be implemented with little additional overhead when compared to traditional approaches

41. Extension of the algorithm • Assign higher weights to Hondas and Toyotas when compared to Teslas, so that the diverse results have more Hondas and Toyotas.

42. Questions? Thank You !