1. Document ranking Paolo Ferragina Dipartimento di Informatica Università di Pisa

2. The big fight: find the best ranking...

3. Ranking: Google vs Google.cn

4. Document ranking Text-based Ranking (1° generation) Reading 6.2 and 6.3

5. Similarity between binary vectors Document is binary vector X,Y in {0,1} D Score: overlap measure What’s wrong ?

6. Normalization Dice coefficient (wrt avg #terms) : Jaccard coefficient (wrt possible terms) : OK, triangular NO, triangular

7. What’s wrong in doc-similarity ? Overlap matching doesn’t consider: Term frequency in a document Talks more of t ? Then t should be weighted more. Term scarcity in collection of commoner than baby bed Length of documents score should be normalized

8. A famous “weight”: tf-idf Number of occurrences of term t in doc d tf t,d where n t = #docs containing term t n = #docs in the indexed collection log n n idf t t         Vector Space model

9. Why distance is a bad idea Sec. 6.3

10. A graphical example Postulate: Documents that are “close together” in the vector space talk about the same things. Euclidean distance sensible to vector length !! t1t1 d2d2 d1d1 d3d3 d4d4 d5d5 t3t3 t2t2  cos(  ) = v  w / ||v|| * ||w|| The user query is a very short doc Easy to Spam Sophisticated algos to find top-k docs for a query Q

11. cosine(query,document) Dot product q i is the tf-idf weight of term i in the query d i is the tf-idf weight of term i in the document cos(q,d) is the cosine similarity of q and d … or, equivalently, the cosine of the angle between q and d. Sec. 6.3

12. Cos for length-normalized vectors For length-normalized vectors, cosine similarity is simply the dot product (or scalar product): for q, d length-normalized.

13. Cosine similarity among 3 docs termSaSPaPWH affection1155820 jealous10711 gossip206 wuthering0038 How similar are the novels SaS: Sense and Sensibility PaP: Pride and Prejudice, and WH: Wuthering Heights? Term frequencies (counts) Note: To simplify this example, we don’t do idf weighting.

14. 3 documents example contd. Log frequency weighting termSaSPaPWH affection3.062.762.30 jealous2.001.852.04 gossip1.3001.78 wuthering002.58 After length normalization termSaSPaPWH affection0.7890.8320.524 jealous0.5150.5550.465 gossip0.33500.405 wuthering000.588 cos(SaS,PaP) ≈ 0.789 × 0.832 + 0.515 × 0.555 + 0.335 × 0.0 + 0.0 × 0.0 ≈ 0.94 cos(SaS,WH) ≈ 0.79 cos(PaP,WH) ≈ 0.69

15. Storage For every term, we store the IDF in memory, in terms of n t, which is actually the length of its posting list (so anyway needed). For every docID d in the posting list of term t, we store its frequency tf t,d which is tipically small and thus stored with unary/gamma. Sec. 7.1.2

16. Vector spaces and other operators Vector space OK for bag-of-words queries Clean metaphor for similar-document queries Not a good combination with operators: Boolean, wild-card, positional, proximity First generation of search engines Invented before “spamming” web search

17. Document ranking Top-k retrieval Reading 7

18. Speed-up top-k retrieval Costly is the computation of the cos Find a set A of contenders, with K < |A| << N Set A does not necessarily contain the top K, but has many docs from among the top K Return the top K docs in A, according to the score The same approach is also used for other (non- cosine) scoring functions Will look at several schemes following this approach Sec. 7.1.1

19. How to select A’s docs Consider docs containing at least one query term. Hence this means… Take this further: 1. Only consider high-idf query terms 2. Champion lists: top scores 3. Only consider docs containing many query terms 4. Fancy hits: for complex ranking functions 5. Clustering Sec. 7.1.2

20. Approach #1: High-idf query terms only For a query such as catcher in the rye Only accumulate scores from catcher and rye Intuition: in and the contribute little to the scores and so don’t alter rank-ordering much Benefit: Postings of low-idf terms have many docs  these (many) docs get eliminated from set A of contenders Sec. 7.1.2

21. Approach #2: Champion Lists Preprocess: Assign to each term, its m best documents Search: If |Q| = q terms, merge their preferred lists (  mq answers). Compute COS between Q and these docs, and choose the top k. Need to pick m>k to work well empirically. Now SE use tf-idf PLUS PageRank (PLUS other weights)

22. Approach #3: Docs containing many query terms For multi-term queries, compute scores for docs containing several of the query terms Say, at least 3 out of 4 Imposes a “soft conjunction” on queries seen on web search engines (early Google) Easy to implement in postings traversal

23. 3 of 4 query terms Brutus Caesar Calpurnia 12358132134 248163264128 1316 Antony 348163264128 32 Scores only computed for docs 8, 16 and 32. Sec. 7.1.2

24. Complex scores Consider a simple total score combining cosine relevance and authority net-score(q,d) = PR(d) + cosine(q,d) Can use some other linear combination than an equal weighting Now we seek the top K docs by net score Sec. 7.1.4

25. Approach #4: Fancy-hits heuristic Preprocess: Assign docID by decreasing PR weight Define FH(t) = m docs for t with highest tf-idf weight Define IL(t) = the rest (i.e. incr docID = decr PR weight) Idea: a document that scores high should be in FH or in the front of IL Search for a t-term query: First FH: Take the common docs of their FH compute the score of these docs and keep the top-k docs. Then IL: scan ILs and check the common docs Compute the score and possibly insert them into the top-k. Stop when M docs have been checked or the PR score becomes smaller than some threshold.

26. Approach #5: Clustering Query LeaderFollower Sec. 7.1.6

27. Cluster pruning: preprocessing Pick  N docs at random: call these leaders For every other doc, pre-compute nearest leader Docs attached to a leader: its followers; Likely: each leader has ~  N followers. Sec. 7.1.6

28. Cluster pruning: query processing Process a query as follows: Given query Q, find its nearest leader L. Seek K nearest docs from among L’s followers. Sec. 7.1.6

29. Why use random sampling Fast Leaders reflect data distribution Sec. 7.1.6

30. General variants Have each follower attached to b 1 =3 (say) nearest leaders. From query, find b 2 =4 (say) nearest leaders and their followers. Can recur on leader/follower construction. Sec. 7.1.6

31. Document ranking Relevance feedback Reading 9

32. Relevance Feedback Relevance feedback: user feedback on relevance of docs in initial set of results User issues a (short, simple) query The user marks some results as relevant or non-relevant. The system computes a better representation of the information need based on feedback. Relevance feedback can go through one or more iterations. Sec. 9.1

33. Rocchio (SMART) Used in practice: D r = set of known relevant doc vectors D nr = set of known irrelevant doc vectors q m = modified query vector; q 0 = original query vector; α, β, γ : weights (hand-chosen or set empirically) New query moves toward relevant documents and away from irrelevant documents Sec. 9.1.1

34. Relevance Feedback: Problems Users are often reluctant to provide explicit feedback It’s often harder to understand why a particular document was retrieved after applying relevance feedback There is no clear evidence that relevance feedback is the “best use” of the user’s time.

35. Pseudo relevance feedback Pseudo-relevance feedback automates the “manual” part of true relevance feedback. Retrieve a list of hits for the user’s query Assume that the top k are relevant. Do relevance feedback (e.g., Rocchio) Works very well on average But can go horribly wrong for some queries. Several iterations can cause query drift. Sec. 9.1.6

36. Query Expansion In relevance feedback, users give additional input (relevant/non-relevant) on documents, which is used to reweight terms in the documents In query expansion, users give additional input (good/bad search term) on words or phrases Sec. 9.2.2

37. How augment the user query? Manual thesaurus (costly to generate) E.g. MedLine: physician, syn: doc, doctor, MD Global Analysis (static; all docs in collection) Automatically derived thesaurus (co-occurrence statistics) Refinements based on query-log mining Common on the web Local Analysis (dynamic) Analysis of documents in result set Sec. 9.2.2

38. Query assist Would you expect such a feature to increase the query volume at a search engine?