1. Search by strategy Arjen P. de Vries arjen@acm.org arjen@acm.org CWI, Spinque, Delft University of Technology

2. Interactive Information Access  Feedback:  Interaction improves information representation  Faceted Browsing:  Interaction can let user take over where machine would fail  Search by Strategy:  Interaction can let user take over where system designer would fail

3. Two extreme views on ‘search’ What is DB?  From business applications  Deductive reasoning  Precise and efficient query processing  Users with technical skills (SQL, XQuery, etc) and precise information needs Selection Books where category=‘CS’ What is IR?  From digital libraries, patent collections, etc  Inductive reasoning  Best-effort processing  Users with low technical skills and imprecise information needs Ranking Books about CS Note: SemWeb more DB than IR!!!

4. Complex search tasks (or, “not-so-simple” search tasks, if you like)  I want to buy a house in Amsterdam and I want it with ‘sfeer’ but still in good shape  I can afford about €350K. I need 3 bedrooms, the size should be about 80m 2. It should have a balcony or a backyard  The closer to the station and an AH, the better. BUT… I do not want to live in Amsterdam- Noord, unless there is a quick bus connection to the ferry  I may be willing to drop some of these constraints, but I’m not sure which

5. Search = IR + DB  Complex search tasks mix exact (DB) and ranked (IR) searches, on structured (DB) and unstructured (IR) data  Current technical solutions support either/or  SQL or Excel for structured data (DB)  Google or Lucene for unstructured data (IR)  Combining results requires significant effort  copy & paste result sets between interfaces, “human (probabilistic) joins”

6. Search = IR on-top-of DB ?  IR on-top-of DB: let exact and ranked operations both be processed by the same engine, so they can be mixed freely  IR responsible for ranking models, using DB as a data-access layer; no physical details necessary  DB responsible for reliable, dynamically optimised, data access; no logical details necessary

7. IR on-top-of DB???  Traditional, general-purpose DB technology cannot compete with custom IR search tools  Working assumption: using column stores should solve the efficiency problem

8. IR on-top-of DB – part I let $c := doc("papers.xml")//DOC[author = "John Smith"] let $q := "//text[about(.//Abstract, IR DB integration)]" for $res in tijah-query($c, $q) return $res/title/text() + + I am an IR and DB expert Hiemstra, Rode, Van Os, Flokstra, OSIR 2006 PF/Tijah: text search in an XML database system

9. IR on-top-of DB DATA AND QUERIES DATA MANAGEMENT Bla bla Bla bla bla, bla bla. Bla. Bla bla Bla bla bla, bla bla. Bla, bla bla Bla bla ENGINEERING Data Model Mismatch!!! To implement IR tasks on top of DB is a tough job! FORMALISM MAPPING Rölleke, Tsikrika, Kazai A general matrix framework for modelling Information Retrieval “IR concepts to matrix spaces” Cornacchia, De Vries, Van Ballegooij, Kersten SRAM (Sparse Relational Array Mapping) “matrix spaces to relational tables”

10. SRAM – query lifecycle

11. IR on-top-of DB – part II let $c := doc("papers.xml")//DOC[author = "John Smith"] let $q := "//text[about(.//Abstract, IR DB integration)]" for $res in tijah-query($c, $q) return $res/title/text() +++ I am an IR expert (needn’t be a DB expert as well) Cornacchia, De Vries, ECIR 2007 A Parametrised Search System

12. SRAM for IR # Language Modelling langmod(Q,d,λ) = sum( [ lm_t(d,t,λ) * Q(t) | t ] ) lm_t(d,t,λ) = log( λ * DTf(d,t) + (1-λ) * Tf(t) ) # Okapi BM25 bm25(Q,d,k1,b) = sum( [ w_t(d,t,k1,b) * Q(t) | t ] ) w_t(d,t,k1,b) = Tidf(t) * (k1+1) * DTf(d,t) / ( DTf(d,t) + k1 * ((1-b) + b * Dnl(d)/avgDl) ) DTnl = mxMult(mxTrnsp(LT), LD) # doc-term freq. Dnl = mvMult(mxTrnsp(LD), L) # doc lenght Tnl = mvMult(mxTrnsp(LT), L) # term freq. DTf = [ DTnl(d,t)/Dnl(d) | d,t ] # doc-term norm. freq. Tf = [ Tnl(t)/nLocs | t ] # term norm. freq. mxTrnsp(A) = [ A(j,i) | i,j ] mxMult(A,B) = [ sum([ A(i,k) * B(k,j) | k ]) | i,j ] mvMult(A,V) = [ sum([ A(i,j) * V(j) | j ]) | i ] Indexing.ram (excerpt) Linear_Algebra.ram (excerpt) Retrieval_Models.ram (excerpt)

13. Parameterised Search System (PSS) Cannot we ‘remove’ this IR engineer from the loop, like DBMS software removes the data engineer from the loop? Cornacchia, De Vries, ECIR 2007 A Parametrised Search System

14. Search by Strategy  Visually construct search strategies by connecting building blocks

15. Strategy Builder

16. From Patent to Inventor

17. Search by Strategy  Visually construct search strategies by connecting building blocks  Each block describes either data or actions upon that data  A search strategy may include multiple data sources  Data sources are internally represented as quadruples, triples extended with an additional probability value  Actions are actually scripts expressed in Fuhr and Roelleke’s PRA (TOIS 1997)

18. 1. Which universities/colleges hold patents? 2. Who are the inventors named in those patents? 3. Which inventors are active in the area of our company? Real-life patent search example: Which researchers associated to universities and colleges should our Human Resources manager know to hire the right people on time?

21. Generate Search Engine

22. Work In Progress…

23. How Strategies Help  Strategies improve communication between search intermediary and user  Encapsulate domain expert knowledge  Abstract representation of search expert knowledge  Analyze information seeking process at any stage  Strategies facilitate knowledge management  Store / share / publish / refine  Strategies mix exact (DB) and ranked (IR) searches  Avoid the need for “human (probabilistic) joins”

24. Implementation  PRA translates into SQL (!)  Current system setup using CWI’s MonetDB column-store  Strategies are dynamically transformed into a REST API

25. Conclusion  Hand over control to the user (or, most likely, the search intermediary)  Patent information specialists  Digital forensics detectives  Librarians / archivists  Real estate agents  Travel agency

26. Open Issues  Assist the user make the best out of their increased level of control  Integrate usage data live system to help improve or adapt strategies  Handle “even larger” scale data  Patent demo fine on ~17GB semi-structured data (i.e., Fairview Research’s Green Energy collection), without specific optimizations, even with fairly large strategies  Close the loop!

27. Current Situation  index ;  repeat {  specify ;  retrieve  } until IR + DB IE

28. Desirable Situation  repeat {  index ;  specify ;  retrieve  } until IR + DB + IE