You are on page 1of 71

Ph.D.

Viva-Voce
VENINGSTON
K
Objectives of
Research
Work

Personalized Information Retrieval System


Using Computational Intelligence Techniques

Introduction

VENINGSTON K

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Senior Research Fellow


Department of Computer Science and Engineering
Government College of Technology, Coimbatore
veningstonk@gct.ac.in
Under the Guidance of
Dr.R.SHANMUGALAKSHMI
Associate Professor, Dept. of CSE, GCT

05 August 2015

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

1 / 71

Ph.D.
Viva-Voce

Presentation Outline

VENINGSTON
K

Objectives of Research Work

Introduction

Literature
Survey

Literature Survey

Proposed
Research
Works

Proposed Research Works


Term Association Graph Model for Document Re-ranking
Topic Model for Document Re-ranking
Genetic Intelligence Model for Document Re-ranking
Swarm Intelligence Model for Search Query Reformulation

Conclusion

References

Publications

Objectives of
Research
Work
Introduction

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

2 / 71

Ph.D.
Viva-Voce

Objectives of Research Work

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

To improve the retrieval effectiveness by employing Term


Association Graph data structure
To enhance a personalized ranking criteria by modeling of
users search interests as topics. Further, employing
Document topic model that integrates User topic model
To realize Genetic Algorithm enabled document re-ranking
scheme
To devise personalized search query suggestion using Ant
Colony Optimization

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

3 / 71

Ph.D.
Viva-Voce

Introduction [1/2]

VENINGSTON
K
Objectives of
Research
Work

Typical Information Retrieval (IR) Architecture

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

4 / 71

Ph.D.
Viva-Voce

Introduction [2/2]

VENINGSTON
K
Objectives of
Research
Work

Why Personalization in Information Retrieval?

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

5 / 71

Ph.D.
Viva-Voce

Classifications of Typical IR systems

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Content-based approach
Simple matching of a query with results - This does not help
users to determine which results are worth

Author-relevancy technique
Citation and hyperlinks - Presents the problem of authoring
bias i.e. results that are valued by authors are not necessarily
those valued by the entire population

Usage rank approach


Actions of users to compute relevancy - Computed from the
frequency, recency, duration of interaction by users

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

6 / 71

Ph.D.
Viva-Voce

Limitations in Typical IR systems

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Most of the techniques measure relevance as a function of


the entire population of users
This does not acknowledge that relevance is relative for
each user
There needs to be a way to take into account that
different people find different things relevant
Users interests and knowledge change over time personal relevance

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

7 / 71

Ph.D.
Viva-Voce

General Approach for mitigating Challenges

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Main ways to personalize a search are Result processing


and Query augmentation

Document Re-ranking
To re-rank the results based upon the frequency, recency, or
duration of usage. Provides users with the ability to identify
the most popular, faddish pages that other users have seen

Query Reformulation
To compare the entered query against the contextual
information available to determine if the query can be
refined/reformulated to include other text

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

8 / 71

Ph.D.
Viva-Voce

General Problem Description

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Diverse interest of search users


Original Query

User 1

User 2

User 3

World cup

football championship

ICC cricket world cup

T20 cricket world cup

India crisis

Economic crisis in India

security crisis in India

job crisis in India

Job search

Student part time jobs

government jobs

Engineering and IT job search

Cancer

astrology and zodiac

lung cancer and prevention

causes of cancer, symptoms and treatment

Ring

Ornament

horror movie

circus ring show

Okapi

animal giraffe

African luxury hand bags

Information retrieval model BM25

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

9 / 71

Ph.D.
Viva-Voce

Literature Survey

VENINGSTON
K
Objectives of
Research
Work

Related work on Re-ranking techniques

Introduction

Paper Title
Implicit preference

Literature
Survey

Hyperlink data

Brin & Page, 1998

Link structure analysis

Computes universal notion of importance

Collaborative filtering

Sarwar et al, 2000

Groupization algorithm

User data are dynamic


Predefined categories are used

Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Author, Year
Sugiyama et al, 2004

Techniques used
Term frequency scheme

Limitations
Noisy browsing history

Categorization

Liu et al, 2004

Mapping queries to related categories

Long-term user behavior

Bennett et al, 2012

Create profiles from entire history

Misses searcher needs for the current task

Short-term user behavior

Cao et al, 2008

Create profiles from recent search session

Lacks in capturing users long term interest

Location awareness

Leung et al, 2010

Location ontology

Captures location information by text matching

Task awareness

Luxenburger et al, 2008

Task language model

Lacks temporal features of user tasks

Tag data

Carman et al, 2008

Content based profiles

Results are biased towards particular user group

Context data

White et al, 2009

User modeling uses Contextual features

Treat all context sources equally

Click data

Liu et al, 2002

Assesses pages frequently clicked

Makes no use of terms and its association

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

10 / 71

Ph.D.
Viva-Voce

Literature Survey

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Related work on Query reformulation techniques


Paper Title
Anchor text

Author, Year
Kraft & Zien, 2004

Techniques used
Mining anchor texts

Limitations
More number of query suggestions

Bipartite graph
Personalized facets

Mei et al, 2008

Prepares morphological different keywords

Individual user intents are not considered

Koren et al, 2008

Employs key-value pair meta-data

Uses frequency based facet ranking

Term association pattern

Wang & Zhai, 2008

Analyze relations of terms inside a query

Click-through data not considered

Rule based classifier

Huang & Efthimiadis, 2009

Matches query with ordered reformulation rules

Semantic associations are missing

Clustering

Jain & Mishne, 2010

Query suggestions are grouped by topics

Drift in user intent to another topic

Merging

Sheldon et al, 2011

Produces results from different reformulations

Random walk on the click graph

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

11 / 71

Ph.D.
Viva-Voce

Proposed Research Works

VENINGSTON
K
Objectives of
Research
Work

Module 1

Introduction

Term Association Graph Model for Document Re-ranking

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Module 2
Topic Model for Document Re-ranking

Module 3
Genetic Intelligence Model for Document Re-ranking

Module 4
Swarm Intelligence Model for Search Query Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

12 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work

1. Term Association Graph Model for Document


Re-ranking
Problem Statement

Introduction
Literature
Survey

How to represent document collection as term graph


model?

Proposed
Research
Works

How to use it for improving search results?

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Methodology
Term graph representation
Ranking semantic association for Re-ranking
TermRank based approach (TRA)
Path Traversal based approach (PTA)
1 PTA1: Naive approach
2 PTA2: Paired similarity document ordering
3 PTA3: Personalized path selection

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

13 / 71

Ph.D.
Viva-Voce

Document Representation

VENINGSTON
K
Objectives of
Research
Work

Sample of OHSUMED (Oregon Health & Science


University MEDline) test Collection

Introduction

DocID

Item-set

Support

Literature
Survey

54711

Ribonuclease, catalytic, lysine, phosphate, enzymatic, ethylation

0.12

Proposed
Research
Works

55199

Ribonuclease, Adx, glucocorticoids, chymotrypsin, mRNA

0.2

62920

Ribonuclease, anticodon, alanine, tRNA

0.1

64711

Cl- channels, catalytic, Monophosphate, cells

0.072

65118

isozyme, enzyme, aldehyde, catalytic

0.096

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

Pn

fd (ti )
Supportd = PN i=1
Pn
j=1
i=1 fd (ti )

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

14 / 71

Ph.D.
Viva-Voce

Term Association Graph Model

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

15 / 71

Ph.D.
Viva-Voce

Ranking Schemes based on Semantic Association

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

16 / 71

Ph.D.
Viva-Voce

Term Rank Approach (TRA)

VENINGSTON
K
Objectives of
Research
Work
Introduction

Rank(ta ) = c

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

tb Ta

Rank(tb )
Ntb

ta and tb are Nodes


Tb is a set of terms ta points to
Ta is a set of terms that point to ta
Ntb = |Tb | is the number of links from ta

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

17 / 71

Ph.D.
Viva-Voce

PTA1: Naive Approach

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

The sequence of documents are chosen from path p3 i.e.


D11 , D1 , D37 , D17 , D22 , andD5 . D11 will be the top ranked
document.

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

18 / 71

Ph.D.
Viva-Voce

PTA2: Paired Similarity Ranking

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works

sim(T1 , T2 ) = 2

depth(LCS)
depth(T1 )+depth(T2 )

T1 and T2 denote the term nodes in Term Association


Graph TG
LCS denote the Least Common Sub-Sumer of T1 and T2
depth(T ) denote the shortest distance from query node q
to a node T on TG

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

19 / 71

Ph.D.
Viva-Voce

PTA3: Personalized Path Selection

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

20 / 71

Ph.D.
Viva-Voce

PTA3: Personalized Path Selection

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

21 / 71

Ph.D.
Viva-Voce

PTA3: Personalized Path Selection

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey

PSCweight

1
=
|t|

Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

""#topics
X
i=1

# 
#
X
#t
/T
(sivi (
t Ti )) 1
|t|

PSCweight is the Personalized Search Context Weight


|t| is the total number of terms in dfspath including
query term
T is the set of user interested topics
sivi is the search interest value of i th topic

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

22 / 71

Ph.D.
Viva-Voce

Experimental Dataset

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

23 / 71

Ph.D.
Viva-Voce

Evaluation Measures

VENINGSTON
K
Objectives of
Research
Work

Subjective Evaluation
1

Information Richness

Introduction

InfoRich(Rm ) =

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

Div (Rm
Nk
X 1 X
1
InfoRich(dki )
Div (Rm )
Nk
k=1

i=1

Objective Evaluation
Precision P = #RelevantRetrived
k
2 Recall P = #RelevantRetrieved
Total#Relevant
1

P|Q|

AvgPrecision(q)

Mean Average Precision MAP = q=1 |Q|


PR
AvgPrecision(q) = R1 k=1 ((P@k) . (rel (k)))
P|Q| 1
4 Mean Reciprocal Rank MRR = 1
i=1 ranki
|Q|
5 Normalized Discounted
Cumulative
Gain
ri
P
3

NDCGk =

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

2 1
k
i=1 log2 (i+1)

IDCGk

Ph.D. Viva-Voce

05 August 2015

24 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Experimental Results & Analysis [1/3]


Non-Personalized Evaluation on Real Dataset

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

25 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Experimental Results & Analysis [2/3]


Non-Personalized Evaluation on Synthetic Dataset

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

26 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Experimental Results & Analysis [3/3]


Personalized Evaluation on Real Dataset

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

27 / 71

Ph.D.
Viva-Voce

Motivation to Module 2

VENINGSTON
K
Objectives of
Research
Work
Introduction

Summary of Module 1
1

Employs term
association graph
model

Suggested different
methods to enhance
the document
re-ranking

Captures hidden
semantic association

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Exploits topical representation


for identifying user interest.
Matching of documents and
queries is not done with topical
representation. To explore topic
model to find relevant
documents by matching topical
features

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

28 / 71

Ph.D.
Viva-Voce

2. Topic Model for Document Re-ranking

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Problem Statement
How to model and represent past search contexts?
How to use it for improving search results?
Methodology
User search context modeling
1
2
3

User profile modeling


Learning user interested topic
Finding document topic

Personalized Re-ranking process


Exploiting user interest profile model
Computing personalized score for document using user
model
3 Generating personalized result set by re-ranking
1
2

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

29 / 71

Ph.D.
Viva-Voce

User search context modeling

VENINGSTON
K
Objectives of
Research
Work

User profile modeling

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

u = UPwi
UPwi History (D) = P(wi ) = P

tfwi ,D
wi D tfwi ,D

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

30 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

User search context modeling


Learning user interested topic

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Finding document topic

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

31 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Personalized Re-ranking process


Exploiting user interest profile model

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

KLD(Td k Tu ) =

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

X
tDU

Ph.D. Viva-Voce

P(Td (t))log

P(Td (t))
P(Tu (t))
05 August 2015

32 / 71

Ph.D.
Viva-Voce

Personalized Re-ranking process

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

Computing personalized score for document using user


model
(D | u )P(Q | D, u )
P(D | Q, u ) =
P(Q | u )
Y
P(Q | D, u ) = P(Q | Td , Tu )+
(P(qi | u )+(1)P(qi | D))
qi Q

P(Q | Tu , Td ) =

(P(qi | Tu ) + (1 )P(qi | Td ))

qi Q

Generating personalized result set by re-ranking


1
2

The documents are scored based on P(Q | D, u )


Result set is re-arranged based on descending order of the
personalized score

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

33 / 71

Ph.D.
Viva-Voce

Experimental Dataset

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

34 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Parameter Tuning
Learning and Parameters

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

35 / 71

Ph.D.
Viva-Voce

Experimental Results & Analysis [1/2]

VENINGSTON
K

Evaluation on Real Dataset


Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

36 / 71

Ph.D.
Viva-Voce

Experimental Results & Analysis [2/2]

VENINGSTON
K

Evaluation on AOL Dataset


Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

37 / 71

Ph.D.
Viva-Voce

Motivation to Module 3

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Summary of Module 2
1

Client side
personalization

Insensitive to the
number of Topics

Not all the queries


would require
personalization to be
performed

Explores topic model for finding


relevant documents using
topical features. To learn a
topic model on a representative
subset of a collection using
Genetic Intelligence technique

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

38 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work

3. Genetic Intelligence Model for Document


Re-ranking
Problem Statement

Introduction

How to represent documents as chromosomes?

Literature
Survey

How to evaluate fitness of search results?

Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

Methodology
Apply GA with an adaptation of probabilistic model
Probabilistic similarity function has been used for fitness
evaluation
Documents are assigned a score based on the probability
of relevance
Probability of relevance are sought using GA approach in
order to optimize the search process i.e. finding of relevant
document not by assessing the entire corpus or collection

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

39 / 71

Ph.D.
Viva-Voce

Why GA for IR?

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

When the document search space represents a high


dimensional space i.e. the size of the document corpus is
multitude in IR
GA is the searching mechanisms known for its quick search
capabilities
When no relevant documents are retrieved in top order
with the initial query
The probabilistic exploration induced by GA allows the
exploration of new areas in the document space.

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

40 / 71

Ph.D.
Viva-Voce

Steps in GA for IR

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

41 / 71

Ph.D.
Viva-Voce

Fitness Evaluation

VENINGSTON
K
Objectives of
Research
Work

Representations of Chromosomes

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion

Probabilistic Fitness Functions


P(q | d) =

| w )P(w | d))
P
2 P(q | d) = P(q | C ) + (1 )
w d (P(q | w )P(w | d))
1

w d (P(q

References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

42 / 71

Ph.D.
Viva-Voce

Selection & Reproduction

VENINGSTON
K
Objectives of
Research
Work

Roulette-wheel selection

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Reproduction Operators
1
2

Crossover
Mutation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

43 / 71

Ph.D.
Viva-Voce

Experimental Dataset

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

44 / 71

Ph.D.
Viva-Voce

Parameter Tuning

VENINGSTON
K
Objectives of
Research
Work
Introduction

Learning Pc and Pm Parameters

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

45 / 71

Ph.D.
Viva-Voce

Experimental Results & Analysis [1/2]

VENINGSTON
K
Objectives of
Research
Work

Evaluation on Benchmark Dataset

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

46 / 71

Ph.D.
Viva-Voce

Experimental Results & Analysis [2/2]

VENINGSTON
K

Evaluation on Real Dataset


Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

47 / 71

Ph.D.
Viva-Voce

Motivation to Module 4

VENINGSTON
K
Objectives of
Research
Work
Introduction

Summary of Module 3
1

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Explored the utility of


incorporating GA to
improve re-ranking

Adaptation of
personalization in GA
provides more desirable
results

Not all the queries


would require
personalization to be
performed

The graph representation of


documents best suit the
application of Swarm
Intelligence model. To simulate
ACO in graph structure based
on behavior of ants seeking a
path between their colony and
source of food for search query
reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

48 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

4. Swarm Intelligence Model for Search Query


Reformulation

Objectives of
Research
Work
Introduction

Problem Statement

Literature
Survey

How to address vocabulary mismatch problem in IR?

Proposed
Research
Works

How to change the original query to form a new query


that would find better relevant documents?

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Methodology
Exploits Ant Colony Optimization (ACO) approach to
suggest related key words
The self-organizing principles which allow the highly
coordinated behavior of real ants that collaborate to solve
computational problems

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

49 / 71

Ph.D.
Viva-Voce

ACO for Query Reformulation

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

Terminologies
Artificial Ant
Pheromone
Typical Ant System

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

50 / 71

Ph.D.
Viva-Voce

ACO Implementation

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

51 / 71

Ph.D.
Viva-Voce

Characteristics of Artificial Ant

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Notion of autocatalytic behavior


Chooses the query term to go with a transition probability
as a function of the similarity i.e. amount of trail present
on the connecting edge between terms
Navigation over retrieved documents for a query is treated
as ant movement over graph
When the user completes a tour, a substance called trail or
trace or pheromone is laid on each edge

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

52 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Similarity
Transition Probability

Objectives of
Research
Work

[ij (t)] [ij ]


pijk (t) = P

k [ik (t)] [ik ]

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

ij is a static similarity weight


ij is a trace deposited by users
Trail Deposition
ij (t + 1) = p ij (t) + ij

p is the rate of trail decay per time interval i.e. pheromone


evaporation factor

Conclusion
References

ij is the sum of deposited trails by users

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

53 / 71

Ph.D.
Viva-Voce

Experimental Dataset

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey

AOL Search query log

Proposed
Research
Works

Only the queries issued by at least 10 users were employed


and the pre-processed documents retrieved for that query
were used to construct graph

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

270 single and two word queries issued by different users


from AOL search log are taken

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

54 / 71

Ph.D.
Viva-Voce

Experimental Evaluation

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Baseline Methods
Association Rule based approach (AR)
SimRank Approach (SR)
Backward Random Walk approach (BRW)
Forward Random Walk approach (FRW)
Traditional ACO based approach (TACO)
Parameter setting
Depth was set as 5 i.e. top ranked 5 related queries
Evaporation factor (p) was set to 0.5

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

55 / 71

Ph.D.
Viva-Voce
VENINGSTON
K

Experimental Results & Analysis


Manual Evaluation

Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Benchmark Evaluation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

56 / 71

Ph.D.
Viva-Voce

Summary

VENINGSTON
K
Objectives of
Research
Work
Introduction

Summary of Module 4

Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Terms in the initial set of documents constitute potential


related terms

Semantically related keywords are suggested to the initial


query

Single word queries are treated as an ambiguous one

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

57 / 71

Ph.D.
Viva-Voce

Conclusion

VENINGSTON
K
Objectives of
Research
Work

Efficient retrieval of Journal articles


The graph structure may signify grammatical relations
between terms

Introduction
Literature
Survey

Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

Usage of Term Association Graph

Integration of Document Topic model and User Topic


model
Effective in general search
This may incorporate live user feedback

GA based document fitness evaluation


Good in document space exploration
Chromosomes representation may be improved

ACO based query reformulation


Exploits collaborative knowledge of users
If the solution is badly chosen, the probability of a bad
performance is high

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

58 / 71

Ph.D.
Viva-Voce

Further Extension

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Efficient updating policy for user interest models

Account individual user specific context for generating


query refinements

Medical Information Retrieval (Eg. PubMed, WebMD,


etc.)

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

59 / 71

Ph.D.
Viva-Voce

Book References

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Salton & McGill (1986)


Introduction to modern information retrieval
McGraw-Hill , New York.
Baeza-Yates & Ribeiro-Neto (1999)
Modern Information Retrieval
Addison Wesley .
Manning et al (2008)
Introduction to Information Retrieval
Cambridge University Press .
Goldberg (1989)
Genetic Algorithms in Search, Optimization, and Machine Learning
Addison-Wesley .

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

60 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

References [1/7]
Matthijis & Radlinski (2012)
Personalizing Web Search using Long Term Browsing History
In Proc. 4th ACM WSDM , 25 34.
Agichtein et al (2006)
Improving Web Search Ranking by Incorporating user behavior
information
In Proc. 29th ACM SIGIR , 19 26.
Ponte & Croft (1998)
A language modeling approach to information retrieval
In Proc. 21st ACM SIGIR , 275 281.
Lafferty et al (2001)
Document language models, query models, and risk minimization for
information retrieval
In Proc. 24th ACM SIGIR , 111 119.
Kushchu (2005)
Web-Based Evolutionary and Adaptive Information Retrieval
IEEE Trans. Evolutionary Computation 9(2), 117 125.

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

61 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

References [2/7]
Leung & Lee (2010)
Deriving Concept-based User profiles from Search Engine Logs
IEEE Trans. Knowledge and Data Engineering 22(7), 969 982.
Blanco & Lioma (2012)
Graph-based term weighting for information retrieval
Springer Information Retrieval 15(1), 54 92.
Dorigo et al (2006)
Ant Colony Optimization
IEEE Computational Intelligence Magazine 1(4), 28 39.
Sugiyama et al (2004)
Adaptive web search based on user pro?le constructed without any
effort from users
In Proc. 13th Intl.Conf. World Wide Web, 675 684.
Brin Page (1998)
The Anatomy of a Large-Scale Hypertextual Web Search Engine
Elsevier Journal on Computer Networks and ISDN Systems 30(1-7),
107 117.

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

62 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

References [3/7]
Eirinaki Vazirgiannis (2005)
UPR:Usage-based page ranking for web persoanalization
In Proc. 5th IEEE Intl. Conf. Data Mining, 130 137.
Sarwar et al (2000)
Analysis of Recommendation Algorithms for E-commerce
In Proc. 2nd ACM Intl. Conf. Electronic commerce, 158 167.
Liu et al (2004)
Personalized web search for improving retrieval effectiveness
IEEE Trans. Knowledge and Data Engineering 16(1), 28 40.
Bennett et al (2012)
Modeling the impact of short- and long-term behavior on search
personalization
In Proc. 35th ACM SIGIR , 185 194.
cao et al (2008)
Context-Aware Query Suggestion by Mining Click-Through
In Proc. 14th ACM SIGKDD , 875 883.

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

63 / 71

Ph.D.
Viva-Voce

References [4/7]

VENINGSTON
K
Objectives of
Research
Work

carman et al (2008)
Tag data and personalized Information Retrieval
In Proc. ACM workshop Search in social media , 27 34.

Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

Leung et al (2010)
Personalized Web Search with Location Preferences
In Proc. 26th IEEE Intl. Conf. Data Engineering , 701 712.
Luxenburger et al (2008)
Task-aware search personalization
In Proc. 31st ACM SIGIR , 721 722.
white et al (2009)
Predicting user interests from contextual information
In Proc. 32nd ACM SIGIR , 363 370.
White et al (2013)
Enhancing personalized search by mining and modeling task behavior
In Proc. 22nd Intl. Conf. World Wide Web , 1411 1420.

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

64 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept.

References [5/7]
Vallet et al (2010)
Personalizing web search with Folksonomy-Based user and document
profiles
In Proc. 32nd European conference on Advances in IR, 420 431.
Jansen et al (2007)
Determining the user intent of web search engine queries
In Proc. 6th Intl. Conf. World Wide Web, 1149 1150.
Jansen et al (2000)
Real life, real users, and real needs: a study and analysis of user
queries on the web
Elsevier Information Processing and Management 36(2), 207 227.
Daoud et al (2008)
Learning user interests for a session-based personalized Search
In Proc. 2nd Intl. symposium on Information interaction in context ,
57 64.
Kraft Zien (2004)
Mining Anchor Text for Query Refinement
In Proc. 13th Intl. Conf.
World Wide Web , 666 674.
of CSE, GCT)
Ph.D. Viva-Voce
05 August 2015

65 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

References [6/7]
Dang Croft (2010)
Query Reformulation Using Anchor Text
In Proc. 3rd ACM WSDM, 41 50.
Mei et al (2008)
Query suggestion using hitting time
In Proc. 17th ACM CIKM , 469 478.
Koren et al (2008)
Personalized interactive faceted search
In Proc. 17th Intl. Conf. World Wide Web, 477 486.
Wang Zhai (2005)
Mining term association patterns from search logs for effective query
Reformulation
In Proc.ACM CIKM , 479 488.
Huang Efthimiadis (2009)
Analyzing and evaluating query reformulation strategies in web search
logs
In Proc. 18th ACM CIKM , 77 86.

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

66 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

References [7/7]
Jain Mishne (2010)
Organizing query completions for web search
In Proc. ACM CIKM , 1169 1178.
Sadikov et al (2010)
Clustering query refinements by user intent
In Proc. 19th Intl. Conf. World Wide Web, 841 850.
Bhatia (2011)
Query suggestions in the absence of query logs
In Proc. 34th ACM SIGIR, 795 804.
Sheldon et al (2011)
LambdaMerge: Merging the results of query reformulations
In Proc. 4th ACM WSDM, 117 125.
Goyal et al (2012)
Query representation through lexical association for information
retrieval
IEEE Trans. Knowledge and Data Engineering 24(12), 2260 2273.

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

67 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept.

Journal Publications
Veningston, & Shanmugalakshmi (2015)
Semantic Association Ranking Schemes for Information Retrieval
Applications using Term Association Graph Representation
Sadhana - Academy Proceedings in Engineering Sciences , Springer
Publication. [Annexure I]
Veningston & Shanmugalakshmi (2014)
Computational Intelligence for Information Retrieval using Genetic
Algorithm
INFORMATION - An International Interdisciplinary Journal 17(8),
3825 3832. [Annexure I]
Veningston & Shanmugalakshmi (2014)
Combining User Interested Topic and Document Topic for
Personalized Information Retrieval
Lecture Notes in Computer Science Springer Publication 8883 , 60
79.[Annexure II]
Veningston & Shanmugalakshmi (2014)
Efficient Implementation of Web Search Query reformulation using
Ant Colony Optimization
of
CSE, GCT)
Ph.D. Viva-Voce
05 August
2015, 80 68
Lecture
Notes in Computer
Science Springer Publication
8883
/ 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References

International Conference Publications [1/2]


Veningston, & Shanmugalakshmi (2015)
Personalized Location aware Recommendation System
In Proc. 2nd IEEE Intl. Conf. Advanced Computing and
Communication Systems , Indexed in IEEE Xplore. [Best Paper]
Veningston & Shanmugalakshmi (2014)
Information Retrieval by Document Re-ranking using Term
Association Graph
In Proc. ACM Intl. Conf. Interdisciplinary Advances in Applied
Computing , Indexed in ACM Digital Library. [Best Paper]
Veningston & Shanmugalakshmi (2014)
Personalized Grouping of User Search Histories for Efficient Web
Search
In Proc. 13th WSEAS Intl. Conf. Applied Computer and Applied
Computational Science , 164 172.
Veningston & Shanmugalakshmi (2013)
Statistical language modeling for personalizing Information Retrieval
In Proc. 1st IEEE Intl. Conf. Advanced Computing and
Communication Systems , Indexed in IEEE Xplore. [Best Paper]

Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

69 / 71

Ph.D.
Viva-Voce

International Conference Publications [2/2]

VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works
Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Veningston, Shanmugalakshmi & Ruksana (2013)


Context aware Personalization for Web Information Retrieval: A Large
scale probabilistic approach
In Proc. Intl. Conf. Cloud and Big Data Analytics , PSG College of
Technology.
Veningston & Shanmugalakshmi (2012)
Enhancing personalized web search Re-ranking algorithm by
incorporating user profile
In Proc. 3rd IEEE Intl. Conf. Computing, Communication and
Networking Technologies , Indexed in IEEE Xplore.

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

70 / 71

Ph.D.
Viva-Voce
VENINGSTON
K
Objectives of
Research
Work
Introduction
Literature
Survey
Proposed
Research
Works

Thank You & Queries

Term
Association
Graph Model for
Document
Re-ranking
Topic Model for
Document
Re-ranking
Genetic
Intelligence
Model for
Document
Re-ranking
Swarm
Intelligence
Model for Search
Query
Reformulation

Conclusion
References
Publications
VENINGSTON K (Dept. of CSE, GCT)

Ph.D. Viva-Voce

05 August 2015

71 / 71

You might also like