Personalized Information Retrieval in Context

1
Personalized Information Retrieval in Context

David Vallet Universidad Autónoma de Madrid,
Escuela Politécnica Superior,Spain
2
Overview

• Motivation
• Ontology-Based Content Retrieval
• Personalization
• Personalization in Context
• Building a Semantic Runtime Context
• Contextual Preference Activation
• Conclusions

3
Motivation

• Requirements of two different multimedia
  applications european research projects digital
  album (aceMedia) and a news service (MESH)

• Indicate users preferences
• Content
• High level Topics
• Low level
• Topic sub-categories
• Geographical area
• Personalised content
• Search results
• Browsing
• Context awareness
• Temporal preference
• Different scopes
• Session focused interests

Ontology-Based Preference Representation
Personalisation in Context
4
Ontology-Based Content Retrieval
Info need
Formal query
Goal Improve keyword-based search
Query engines Inference engines
Ontology KB
Annotation
Documents
Search space
Returned documents
5
Ontology-Based Content Retrieval
q d1 d2
x1 q1 d11 d21
x2 q2 d12 d22
x3 q3 d13 d23
6
Personalization
Preferences/Context
Annotation
Documents
Search space
7
Personalization
Personalization effect
x2
x1
x1, x2, x3 domain ontology O
x3
8
Personalization
Ontology-Based Preference Representation

• Concepts VS Keywords
• Interoperability
• Precision
• Hierarchical Representation
• Inference

9
Personalization
Ontology-Based Preference Representation
Geographical Region
C
Islands
C
Leisure
visit
Spanish Islands
C
C
Travel
Canada
C
USA Islands
C
C
Island Travel
USA
C
C
Hawaii
I
locatedIn
Movies
Florida
C
C
Music
C
Techno
C
Classical
C
Pop
C
Hawaii Tourist Guide
10
Personalisation in Context

• Combination of long-term (preferences)
  short-term (context) user interests and needs
• Not all user preferences are relevant all the
  time which ones?
• Partial answer focus on current semantic
  context, discard out of context ones
• Notion of context
• Defined as the set of background themes under
  which user activities occur within a given unit
  of time
• Represented as a set of weighted ontology
  concepts involved in user actions within a
  session
• Captured?
• Build a runtime context extracting concepts
  from queries and documents selected by the user
• Used?
• Contextual preference activation Analyze
  semantic connections between preference and
  context concepts
• Personalization retrieval in context Filter user
  preferences, only those related to the context
  are activated

11
Building a Runtime Context
Concepts, t
Action Query
Contextt
11
12
Contextual Preference Activation

• preference for x px
• r (x,y)

? preference for y px w (r)
Constrained Spreading Activation
py 0.4 0.8 ? 0.5
px 0.8
py 0.724 0.4 (1 - 0.4) ? 0.9 ? 0.6
w (r) 0.5
C
C
nextTo r
Beach x
Sea y
13
Contextual Preference Activation
Domain concepts
14
Personalization in Context
x2
x1
x1, x2, x3 domain ontology O
x3
15
Conclusions

• Semantic concepts VS plain terms
• Exploitation of semantic relation
• Semantic runtime context
• Context Filtering of user preference

16
References

• Semantic Search
• P. Castells, M. Fernández, and D. Vallet. An
  Adaptation of the Vector-Space Model for
  Ontology-Based Information Retrieval. IEEE
  Transactions on Knowledge and Data Engineering,
  2007. In press.
• Personalization
• D. Vallet, P. Mylonas, M. A. Corella, J. M.
  Fuentes, P. Castells, and Y. Avrithis. A
  Semantically-Enhanced Personalization Framework
  for Knowledge-Driven Media Services. IADIS
  WWW/Internet Conference (ICWI 2005). Lisbon,
  Portugal, October 2005.
• Personalization in context
• D. Vallet, M. Fernández, P. Castells, P. Mylonas,
  and Y. Avrithis. Personalized Information
  Retrieval in Context. 3rd International Workshop
  on Modeling and Retrieval of Context (MRC 2006)
  at the 21st National Conference on Artificial
  Intelligence (AAAI 2006). Boston, USA, July 2006.
  
• Ranking Aggregation
• M. Fernndez, D. Vallet, and P. Castells. Using
  Historical Data to Enhance Rank Aggregation. 29th
  Annual International ACM Conference on Research
  and Development on Information Retrieval (SIGIR
  2006), Poster Session. Seattle, WA, August 2006.
• Tuning Personalization
• P. Castells, M. Fernndez, D. Vallet, P. Mylonas,
  and Y. Avrithis. Self-Tuning Personalized
  Information Retrieval in an Ontology-Based
  Framework. 1st IFIP WG 2.12 WG 12.4
  International Workshop on Web Semantics (SWWS
  2005), November 2005. Springer Verlag Lecture
  Notes in Computer Science, Vol. 3762. Meersman,
  R. Tari, Z. Herrero, P. (Eds.), 2005, ISBN
  3-540-29739-1, pp. 977-986.

17
Thank You!