Polyrepresentation in Complex (Book) Search
Tasks
How can we use what the others said?
Ingo Frommholz
University of Bedfor...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Outline
Motivation
Abstraction for Complex...
Motivation
. . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Motivating Example: Book Store
“Good intro...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Motivating Example: Book Store
“Good intro...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
IN Facets and Polyrepresentation
“Good int...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Polyrepresentation
Book Store Scenario
Con...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Another Challenge
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Some Approaches
▶ POLAR – abstraction laye...
Abstraction for Complex Search Tasks –
POLAR
. . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Abstraction for Information Retrieval
▶ Pr...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Annotation Model: Classes, Properties
[Fro...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
POLAR Motivation
Utilising structured anno...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
POLAR
Probabilistic Object-Oriented Logics...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Structured Documents and Content Level Ann...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Meta Level Annotations
▶ d[ p @j] : d anno...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Positive and Negative Annotations
▶ d[ p -...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Fragments
▶ A fragment f of a document d t...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
References
a[ p =>o] : a references an obj...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Attributes and Classifications
▶ Attributes...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Database and Structure-oriented Queries
Fa...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Content-oriented Queries
▶ All documents a...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
No Augmentation
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
With Knowledge Augmentation
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Example Application: Ratings
d1[ 0.7 datab...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Annotation-based Trustworthiness/Belief
0....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Implementation: POLAR Execution/Translatio...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Implementation: POLAR → FVPD
POLAR
d1[ 0.6...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Implementation: FVPD → pDatalog
FVPD
0.6 t...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Implementation: pDatalog → PRA
pDatalog
0....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
POLAR Evaluation
▶ Evaluation of knowledge...
Quantum-inspired Information Access
. . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Quantum-inspired Information Access
Quantu...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
User Interaction and Feedback
R∗
|ϕ1
|ϕ2
|...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
User Interaction and Feedback
R∗
|ϕ1
|ϕ2
|...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Polyrepresentation/Multiple Evidence
[From...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Properties of the Framework
▶ Each user in...
Polyrepresentation and Clustering
. . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Polyrepresentation and Clustering
▶ Polyre...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Polyrepresentation and Clustering
▶ Mappin...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Information Need-based Vector
▶ Let REPin ...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Document-based Polyrepresentation Vector
▶...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Some Findings (using iSearch)
▶ Some stati...
Conclusion
. . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Conclusion
▶ The rich source of evidence i...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Thanks for your attention!
Questions?
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography I
Abbasi, M. K. and Frommholz...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography II
Digital Libraries (ECDL 20...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography III
Frommholz, I. and Fuhr, N...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography IV
Frommholz, I., Piwowarski,...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography V
Koolen, M. (2014).
"User Re...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Bibliography VI
Piwowarski, B., Frommholz,...
of 52

Polyrepresentation in Complex (Book) Search Tasks - How can we use what the others said?

The task definition of the Social Book Search Lab describes complex goal-oriented and non-goal tasks. To satisfy the resulting information needs, the user can utilise and combine different sources of evidence, like, for instance, metadata (e.g. abstract, title, author) and reviews and ratings provided by the user. The challenge is to support the user in this endeavour to create an effective search experience. To this end, in this talk I will discuss how this challenge relates to the well-known principle of polyrepresentation. I will then introduce a probabilistic logic-based framework called POLAR, which is capable of handling complex queries based on the graph induced by user-generated content. Subsequently I will provide a brief outlook on further formal models that try to support the user beyond the typical query-and-result paradigm. The first one is based on quantum probabilities, neatly combining geometry and probability theory to support different forms of user interaction and polyrepresentation. The latter one combines polyrepresentation with probabilistic clustering and the idea of a simulated user.
Published on: Mar 4, 2016
Published in: Data & Analytics      
Source: www.slideshare.net


Transcripts - Polyrepresentation in Complex (Book) Search Tasks - How can we use what the others said?

  • 1. Polyrepresentation in Complex (Book) Search Tasks How can we use what the others said? Ingo Frommholz University of Bedfordshire ingo.frommholz@beds.ac.uk Twitter: @iFromm CLEF 2015 Social Book Search Workshop September 10, 2015 . . . . . . . . . . . . . . . . . . . .
  • 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outline Motivation Abstraction for Complex Search Tasks – POLAR Quantum-inspired Information Access Polyrepresentation and Clustering Conclusion
  • 3. Motivation . . . . . . . . . . . . . . . . . . . .
  • 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Motivating Example: Book Store “Good introduction to quantum mechanics”
  • 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Motivating Example: Book Store “Good introduction to quantum mechanics”
  • 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IN Facets and Polyrepresentation “Good introduction to quantum mechanics” ▶ Relevance decision goes beyond topicality ▶ Collections like Amazon/LT/BritishLibrary ▶ Rich pool of potentially useful information (metadata, user-generated content) ▶ Different views on documents, relevant for different aspects of the information need (IN) ▶ Combine the evidence (e.g. metadata and user-generated content) to get a more accurate estimation of relevance/usefulness ▶ [Koolen, 2014] puts user-generated content into the index – it worked! ▶ Reviews and tags complimentary to each other and to professional metadata ▶ Polyrepresentation a key principle (exploits different contexts [Ingwersen and Järvelin, 2005])
  • 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polyrepresentation Book Store Scenario Content Author Ratings Comments
  • 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Another Challenge
  • 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Some Approaches ▶ POLAR – abstraction layer for complex search tasks utilising annotations ▶ Quantum Information Access – modelling polyrepresentation and user interaction ▶ Polyrepresentative clustering – supporting different access modes (browsing)
  • 10. Abstraction for Complex Search Tasks – POLAR . . . . . . . . . . . . . . . . . . . .
  • 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Abstraction for Information Retrieval ▶ Provide a task-oriented solution for knowledge engineers ▶ Should not have to bother with the underlying retrieval model/data sources/data storage and organization ▶ Instead focus on the task at hand ▶ Support complex retrieval strategies and information needs ▶ Allows for exploiting task-crossovers and synergies as well as reusing concepts defined for similar tasks
  • 12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Annotation Model: Classes, Properties [Frommholz and Fuhr, 2006b, Agosti et al., 2004]
  • 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . POLAR Motivation Utilising structured annotation hypertexts ▶ Indexing and modelling of structured annotation hypertexts ▶ Querying structured annotation hypertexts ▶ Annotation-based document and discussion search ▶ Support different types of (complex) information needs
  • 14. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . POLAR Probabilistic Object-Oriented Logics for Annotation-based Retrieval [Frommholz and Fuhr, 2006a] ▶ Object-oriented ▶ Classes, instances and relations (attributes), aggregation ▶ Logics ▶ Four-valued logics (true, false, inconsistent, unknown) ▶ Probabilistic ▶ Probabilistic inference and evaluation of rules ▶ Annotation-based retrieval ▶ Models and utilises structured annotation hypertexts ▶ Possible world semantics
  • 15. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Structured Documents and Content Level Annotations ▶ d[ p *a] : d annotated by content annotation a ▶ p access probability
  • 16. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Meta Level Annotations ▶ d[ p @j] : d annotated by meta annotation j
  • 17. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Positive and Negative Annotations ▶ d[ p -*a] : a is negative content annotation ▶ d[ p -@a] : a is negative meta annotation
  • 18. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fragments ▶ A fragment f of a document d that is annotated (highlighted) by a: d[ p1 f|| ... p2 *a|| ]
  • 19. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References a[ p =>o] : a references an object o
  • 20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Attributes and Classifications ▶ Attributes: Turner is the author of a1: a1.author(turner) ▶ Classifications: Tweety is a bird, but Roger Rabbit isn’t: bird(tweety) !bird(roger_rabbit)
  • 21. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Database and Structure-oriented Queries Factual (database-like) queries to the knowledge base. Example: ▶ All annotations written by “turner”: ?- A.author(turner) & annotation(A) Structure-oriented queries to the knowledge base. Examples: ▶ All content level annotations annotating d1: ?- d1[ *A ] ▶ All documents annotated by a1 ?- D[ *a1 ] & document(D)
  • 22. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Content-oriented Queries ▶ All documents about ‘information’ and ‘retrieval’ which are good introductions: ?- document(D) & D[ information & retrieval & @A] & A[ good & introduction ] ▶ All documents having a highlighted part about ‘information’ and ‘retrieval’: retrieve(D) :- document(D) & D[ ||F ] & F|| information & retrieval || ?- retrieve(D)
  • 23. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . No Augmentation
  • 24. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . With Knowledge Augmentation
  • 25. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Example Application: Ratings d1[ 0.7 databases 0.5 @a1 0.5 @a2 ] d2[ 0.8 databases 0.5 @a3 0.5 @a4 ] a1[ 0.8 excellent ] a2[ 0.8 excellent ] a3[ 0.4 excellent ] a4[ 0.2 excellent ] excellent_paper(D) :- D[@A] & A[excellent] ?- D[databases] & excellent_paper(D) 0.49 (d1) 0.224 (d2) databases databases 0.2 0.4 0.8 1 excellent d1 d2 a3 a4 a1 a2
  • 26. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Annotation-based Trustworthiness/Belief 0.7 football a1[ 0.7 football 0.5 -*a3 0.5 -*a4 ] a2[ 0.5 football 0.5 *a5 0.5 *a6 ] topical_relevant(O) :- O[football] 0.6 unconditional_trust(a1) 0.6 unconditional_trust(a2) trustworthy(O) :- unconditional_trust(O) trustworthy(O) :- O[*A] /* positive evidence */ !trustworthy(O) :- O[-*A] /* negative evidence */ relevant(O) :- topical_relevant(O) & trustworthy(O) ?- relevant(O) 0.315 (a2) 0.0735 (a1)
  • 27. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implementation: POLAR Execution/Translation Pipe ▶ Abstraction layer on top of Four-valued Probabilistic Datalog (FVPD) ▶ Implemented in Java ▶ POLAR programs translated into FVPD ▶ Uses HySpirit as FVPD implementation ▶ POLAR programs executed by HySpirit POLAR FVPD PDatalog PRA HySpirit
  • 28. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implementation: POLAR → FVPD POLAR d1[ 0.6 soccer 0.8 s1[ 0.3 music ] 0.7 *a1] a1[ 0.5 football ] document(d1) annotation(a1) rel(D) :- D[*A] & A[football] ?- rel(D) FVPD 0.6 term(soccer,d1). 0.8 acc_subpart(d1,s1). 0.7 acc_canno(d1,a1). 0.3 term(music,s1). 0.5 term(football,a1). instance_of(d1,document,db). instance_of(a1,annotation,db). instance_of(D,rel,db):- acc_canno(D,A) & term(football,A) ?- instance_of(D,rel,db).
  • 29. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implementation: FVPD → pDatalog FVPD 0.6 term(soccer,d1). 0.8 acc_subpart(d1,s1). 0.7 acc_canno(d1,a1). 0.3 term(music,s1). 0.5 term(football,a1). instance_of(d1,document,db). instance_of(a1,annotation,db). instance_of(D,rel,db):- acc_canno(D,A) & term(football,A) ?- instance_of(D,rel,db). pDatalog 0.6 term4(t,soccer,d1) 0.8 acc_subpart4(t,d1,s1) 0.7 acc_canno4(t,d1,a1) 0.3 term4(t,music,s1) 0.5 term4(t,football,a1) instance_of4(t,d1,document,db) instance_of4(t,a1,annotation,db) pos_instance_of(D,’rel’,’db’) :- pos_acc_canno(D,A) & !neg_acc_canno(D,A) & pos_term(football,A) & !neg_term(football,A) ?- pos_instance_of(D,rel,db) & !neg_instance_of(D,rel,db)
  • 30. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Implementation: pDatalog → PRA pDatalog 0.6 term4(t,soccer,d1) 0.8 acc_subpart4(t,d1,s1) 0.7 acc_canno4(t,d1,a1) 0.3 term4(t,music,s1) 0.5 term4(t,football,a1) instance_of4(t,d1,document,db) instance_of4(t,a1,annotation,db) pos_instance_of(D,’rel’,’db’) :- pos_acc_canno(D,A) & !neg_acc_canno(D,A) & pos_term(football,A) & !neg_term(football,A) ?- pos_instance_of(D,rel,db) & !neg_instance_of(D,rel,db) PRA 0.6 term4(t,soccer,d1) 0.8 acc_subpart4(t,d1,s1) ... pos_instance_of = UNITE(pos_instance_of, PROJECT[$1,rel,db] (JOIN[$2=$1]( SUBTRACT(PROJECT[$1,$2] (JOIN[$2=$2](pos_acc_canno, SELECT[$1=football](pos_term))), neg_acc_canno), ... ?- PROJECT[$1](SUBTRACT( PROJECT[$1](...
  • 31. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . POLAR Evaluation ▶ Evaluation of knowledge augmentation approach...can annotations improve retrieval effectiveness? ▶ Discussion search ▶ Annotation view on email messages (W3C discussions) ▶ Knowledge augmentation with annotation targets, fragments and direct annotations ▶ Document search ▶ Using annotations as document context (ZDNet) ▶ Knowledge augmentation (full and radius-1) with annotations (comments) ▶ Significant improvements observed, but some combinations led to significantly worse results
  • 32. Quantum-inspired Information Access . . . . . . . . . . . . . . . . . . . .
  • 33. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quantum-inspired Information Access Quantum Probabilities [van Rijsbergen, 2004, Piwowarski et al., 2010a] R p1 p2 p4 p3 p5 ▶ System uncertain about user’s IN ▶ Expressed by an ensemble S of possible IN vectors : S = {(p1,|φ1 ⟩),...,(pn,|φn ⟩)} ▶ Probability of relevance: Pr(R|d,S) = ∑ i pi ·Pr(R|d,φi ) =||R|φ ⟩||2
  • 34. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . User Interaction and Feedback R∗ |ϕ1 |ϕ2 |ϕ5 |ϕ3 ▶ Outcome of feedback: Query, relevant document, ... ▶ Expressed as subspace ▶ Project IN vectors onto document subspace
  • 35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . User Interaction and Feedback R∗ |ϕ1 |ϕ2 |ϕ4 |ϕ3 |ϕ5 ▶ Outcome of feedback: Query, relevant document, ... ▶ Expressed as subspace ▶ Project IN vectors onto document subspace ▶ Document now gets probability 1 ▶ System’s uncertainty decreases ▶ Also reflects changes in information needs
  • 36. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polyrepresentation/Multiple Evidence [Frommholz et al., 2010] Content Author Ratings Comments ▶ Polyrepresentation space as tensor product of single spaces ▶ Probability that document is in total cognitive overlap: Prpolyrep = Prcontent ×Prratings ×Prauthor ×Prcomments ▶ User interaction may lead us into an entangled state (so far unexplored relationship between polyrepresentation and entanglement)
  • 37. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Properties of the Framework ▶ Each user interaction triggers an observation and thus a change of state ▶ Our evaluation shows that the framework can compete with standard models in ad hoc IR tasks ▶ Different IR tasks can be formulated in this framework (filtering [Piwowarski et al., 2010b], query sessions [Frommholz et al., 2011], summarisation [Piwowarski et al., 2012])
  • 38. Polyrepresentation and Clustering . . . . . . . . . . . . . . . . . . . .
  • 39. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polyrepresentation and Clustering ▶ Polyrepresentation creates partitions ▶ Clustering partitions document sets too ▶ Can clustering help in creating polyrepresentative partitions? ▶ Polyrepresentation Cluster Hypothesis: “documents relevant to the same representations should appear in the same clus- ter” [Frommholz and Abbasi, 2014].
  • 40. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polyrepresentation and Clustering ▶ Mapping of clusters to polyrepresentation (using iSearch [Lykke et al., 2010]) ▶ Simulated user – search strategy: 1. User investigates total cognitive overlap cluster 2. User jumps to different cluster based on preferences 3. The user simulation creates a ranked list of documents
  • 41. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Information Need-based Vector ▶ Let REPin be the set of representations1 of an information need in ▶ Motivated by the Optimum Clustering Framework (OCF), which is based on the probability of relevance [Fuhr et al., 2011] ▶ Pr(R|d,ri ) is computed for each document d and ri ∈ REPin ⃗τin(d) =    Pr(R|d,r1) ... Pr(R|d,rn)    (1) 1 search terms, work task, ideal answer, current info need, background knowledge
  • 42. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Document-based Polyrepresentation Vector ▶ REPd consists of the different representations2 rdi of a document d ▶ Pr(R|rdi ,q) for q (search terms in this case) is computed ⃗τdoc(d) =    Pr(R|rd1,q) ... Pr(R|rdn,q)    (2) 2 title, abstract, body, bibliographic context, references
  • 43. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Some Findings (using iSearch) ▶ Some statistically significant improvements over a BM25 baseline (NDCG@30) using the ranking created by a simple simulated user strategy when concatenating the IN and Document representations [Abbasi and Frommholz, 2015b] ▶ Statistical significant improvements (NDCG) when using document and IN representations separately and assuming an ideal (oracle-based) cluster ranking [Abbasi and Frommholz, 2015a] ▶ This shows us our idea is basically promising! ▶ Finding the total cognitve overlap (TOC) using cluster ranking is challenging [Frommholz and Abbasi, 2014] ▶ Different interpretations of the TOC: The one with the highest precision? The one with the highest pairwise precision? The one where all representations get a high value? ▶ The latter one could be identified more easily (MRR = 0.575 compared to around 0.3 for the others)
  • 44. Conclusion . . . . . . . . . . . . . . . . . . . .
  • 45. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion ▶ The rich source of evidence in SBS should be combined to tackle complex information needs ▶ Probabilistic models for expressing complex information needs and interactive search ▶ POLAR (abstraction for annotation-based search) ▶ Quantum Information Access ▶ Probabilistic polyrepresentative clustering (simulated user) ▶ It seems polyrepresentation can successfully be applied ▶ Good idea to integrate different sources ▶ Need to do it wisely
  • 46. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Thanks for your attention! Questions?
  • 47. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography I Abbasi, M. K. and Frommholz, I. (2015a). Cluster-based Polyrepresentation as Science Modelling Approach for Information Retrieval. Scientometrics, 102(3):2301–2322. Abbasi, M. K. and Frommholz, I. (2015b). Polyrepresentative Clustering: A Study of Simulated User Strategies and Representations. In Mayr, P., Frommholz, I., and Mutschke, P., editors, Proc. of the 2nd Workshop on Bibliometric-enhanced Information Retrieval (BIR2015), pages 47–54, Vienna, Austria. CEUR-WS.org. Agosti, M., Ferro, N., Frommholz, I., and Thiel, U. (2004). Annotations in Digital Libraries and Collaboratories – Facets, Models and Usage. In Heery, R. and Lyon, L., editors, Research and Advanced Technology for Digital Libraries. Proc. European Conference on
  • 48. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography II Digital Libraries (ECDL 2004), Lecture Notes in Computer Science, pages 244–255, Heidelberg et al. Springer. Frommholz, I. and Abbasi, M. K. (2014). On Clustering and Polyrepresentation. In de Rijke, M., Kenter, T., de Vries, A. P., Zhai, C., de Jong, F., Radinsky, K., and Hofmann, K., editors, Proceedings of the European Conference on Information Retrieval (ECIR 2014), volume 1, pages 618–623. Springer. Frommholz, I. and Fuhr, N. (2006a). Evaluation of Relevance and Knowledge Augmentation in Discussion Search. In Gonzalo, J., Thanos, C., Verdejo, M. F., and Carrasco, R. C., editors, Research and Advanced Technology for Digital Libraries. Proc. of the 10th European Conference on Digital Libraries (ECDL 2006), Lecture Notes in Computer Science, pages 279–290, Heidelberg et al. Springer.
  • 49. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography III Frommholz, I. and Fuhr, N. (2006b). Probabilistic, Object-oriented Logics for Annotation-based Retrieval in Digital Libraries. In Nelson, M., Marshall, C., and Marchionini, G., editors, Proc. of the 6th ACM/IEEE Joint Conference on Digital Libraries (JCDL 2006), pages 55–64, New York. ACM. Frommholz, I., Larsen, B., Piwowarski, B., Lalmas, M., Ingwersen, P., and van Rijsbergen, K. (2010). Supporting Polyrepresentation in a Quantum-inspired Geometrical Retrieval Framework. In Proceedings of the 2010 Information Interaction in Context Symposium, pages 115–124, New Brunswick. ACM.
  • 50. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography IV Frommholz, I., Piwowarski, B., Lalmas, M., and van Rijsbergen, K. (2011). Processing Queries in Session in a Quantum-Inspired IR Framework. In Clough, P., Foley, C., Gurrin, C., Jones, G. J. F., Kraaij, W., Lee, H., and Mudoch, V., editors, Proceedings ECIR 2011, volume 6611 of Lecture Notes in Computer Science, pages 751–754. Springer. Fuhr, N., Lechtenfeld, M., Stein, B., and Gollub, T. (2011). The Optimum Clustering Framework : Implementing the Cluster Hypothesis. Information Retrieval, 14. Ingwersen, P. and Järvelin, K. (2005). The turn: integration of information seeking and retrieval in context. Springer-Verlag New York, Inc., Secaucus, NJ, USA.
  • 51. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography V Koolen, M. (2014). "User Reviews in the Search Index? That’ll Never Work!". In Proceedings ECIR 2014, pages 323–334. Lykke, M., Larsen, B., Lund, H., and Ingwersen, P. (2010). Developing a Test Collection for the Evaluation of Integrated Search. In Proceedings ECIR 2010, pages 627–630. Piwowarski, B., Amini, M.-R., and Lalmas, M. (2012). On using a Quantum Physics formalism for Multi-document Summarisation. Journal of the American Society for Information Science and Technology (JASIST).
  • 52. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography VI Piwowarski, B., Frommholz, I., Lalmas, M., and Van Rijsbergen, K. (2010a). What can Quantum Theory Bring to Information Retrieval? In Proc. 19th International Conference on Information and Knowledge Management, pages 59–68. Piwowarski, B., Frommholz, I., Moshfeghi, Y., Lalmas, M., and van Rijsbergen, K. (2010b). Filtering documents with subspaces. In Proceedings of the 32nd European Conference on Information Retrieval (ECIR 2010), pages 615–618. van Rijsbergen, C. J. (2004). The Geometry of Information Retrieval. Cambridge University Press, New York, NY, USA.

Related Documents