SlideShare a Scribd company logo
1 of 32
Download to read offline
Learning to Reinforce
Search Effectiveness
Jiyun Luo, Xuchu Dong, Grace Hui Yang
Georgetown University
ICTIR 2015
1
Search by ‘Test-the-Water’
2
But you are not alone:
Search with a Partner
• A teamwork
• Share a common goal
• find relevant documents
• satisfy long term goal
• Equal partners
• not just being an assistant to
the user
• but also providing influence
• Cooperative exploration
3
Key Idea: Cooperative Exporation
• The two parties
• They talk and they listen
• keep exchanging their ideas
• Take turns to lead the search to the direction
each of them would like this collaboration to go
• Also considering the other’s opinion
4
t=1 Query q1=“hydropower efficiency”
Messages:Messages:
See my new query. Let’s
explore!
5
Example is from TREC 2014 Session 52
t=1 Query q1=“hydropower efficiency”
Retrieved docs D1 “…renewable energy…”
Messages:Messages:
Check it out! Documents
I’ve ranked high are
relevant
See my new query. Let’s
explore!
6
Example is from TREC 2014 Session 52
t=1
t=2
Query q1=“hydropower efficiency”
Clicked d2 in D1
Query q2=“hydropower environment”
Retrieved docs D1 “…renewable energy…”
Messages:Messages:
Check it out! Documents
I’ve ranked high are
relevant
See my new query. Let’s
explore!
Documents I’ve clicked
look relevant!
My new query is on
another subtopic. Let’s
explore
7
Example is from TREC 2014 Session 52
t=1
t=2
Retrieved docs D2
Query q1=“hydropower efficiency”
Clicked d2 in D1
Query q2=“hydropower environment”
Retrieved docs D1 “…renewable energy…”
Messages:Messages:
Check it out! Documents
I’ve ranked high are
relevant
Check it out! Documents
I’ve ranked high are
relevant.
See my new query. Let’s
explore!
Documents I’ve clicked
look relevant!
My new query is on
another subtopic. Let’s
explore
8
Example is from TREC 2014 Session 52
t=1
t=2
t=3
Retrieved docs D2
Query q1=“hydropower efficiency”
Clicked d2 in D1
Query q2=“hydropower environment”
Clicked d2 in D2
Query q3=“hydropower damage”
Retrieved docs D1 “…renewable energy…”
Messages:Messages:
Check it out! Documents
I’ve ranked high are
relevant
Check it out! Documents
I’ve ranked high are
relevant.
See my new query. Let’s
explore!
Documents I’ve clicked
look relevant!
Documents I’ve
clicked look relevant!
My new query is on
another subtopic. Let’s
explore
My new query is still on
the same subtopic. Let’s
find out more about it.
9
Example is from TREC 2014 Session 52
t=1
t=2
t=3
…
t=4
Retrieved docs D2
Retrieved docs D3
Query q1=“hydropower efficiency”
Clicked d2 in D1
Query q2=“hydropower environment”
Clicked d2 in D2
Query q3=“hydropower damage”
Retrieved docs D1 “…renewable energy…”
Messages:Messages:
Check it out! Documents
I’ve ranked high are
relevant
Check it out! Documents
I’ve ranked high are
relevant.
See my new query. Let’s
explore!
Documents I’ve clicked
look relevant!
Documents I’ve
clicked look relevant!
Want to explore? I’ve
diversified my results.
My new query is on
another subtopic. Let’s
explore
My new query is still on
the same subtopic. Let’s
find out more about it.
10
Example is from TREC 2014 Session 52
Opinions about Two Things
• Relevance
• Which documents (that
you have just marked/
retrieved/recommended)
are relevant
• Desire of Exploration
• How exploratory I want
us to be, as a team
11
How to Express the Opinions/
Feedback
• Relevance is “demonstrated by examples”:
• Query is a piece of short text sent from the user
• Clicked snippets/documents are long pieces of
text sent from user
• Documents are long text sent from the search
engine
• Desire of Exploration is shown by
• Query changes
• Diversified results
12
A Contextual Bandit Formulation
of a Decision-Making Distribution
P(relevant) = 1 ✏ P(explore) = µ
P(J = RE|o, a, ⇡⇤
) = (1 ✏)µ
P(J = NRE|o, a, ⇡⇤
) = ✏µ
P(J |o, a, ⇡⇤
) = P(relevant)P(explore)
P(J = RNE|o, a, ⇡⇤
) = (1 ✏)(1 µ)
P(J = NRNE|o, a, ⇡⇤
) = ✏(1 µ)
13
Relevance Feedback from
the User
• 1 SAT-Clicked out of 10 retrieved,
✏ = 1
# of SAT-Clicked documents 2 Dt 1
# of returned documents 2 Dt 1
14
" = 1
1
10
= 0.9
smoking quitting!q2! hypnosis !
Rank 1: Easy Ways to Quit Smoking | Quit Smoking Help … !
…!
Rank 3: Quit Smoking Toolbox - Quit Smoking - Nicotine Addiction …!
…!
Rank 6: Quit Smoking Hypnosis, Stop Smoking Hypnosis CDs…!
…!
Rank 10: …!
SAT-Clicked. !
Dwell time: 40 seconds!
D1!
Exploration Feedback from
the User
• 1 query change , 3 terms in the new query in
total
µ = 1
# of query changes 2 Dt 1
# of permutations of query terms 2 Dt 1
15
smoking quitting!q2! hypnosis !
+∆q"
Rank 1: Easy Ways to Quit Smoking | Quit Smoking Help … !
…!
Rank 3: Quit Smoking Toolbox - Quit Smoking - Nicotine Addiction …!
…!
Rank 6: Quit Smoking Hypnosis, Stop Smoking Hypnosis CDs…!
…!
Rank 10: …!
D1!
Query reformulation using words in
previous search results!
2 Dt 1
µ = 1
1
3!
= 0.83
Relevance Feedback from
the Search Engine
• Highly scored documents
• Needs consistency in ranking scores
• Could be hard to get
• Highly ranked documents
✏ = 1
# of relevant documents 2 top n retrieved
n
16
Relevance Feedback from
the Search Engine
17
smoking quitting!q2!
D2! Rank 1: Quit Smoking Hypnosis | Stop Smoking Hypnosis CDs Quit Smoking
Hypnosis Neuro…!
…!
Rank 4: Quit Smoking with Video Hypnosis Home Shopping Cart…!
…!
Rank 10: …!
hypnosis !
• 8 out of 10 top retrieved documents are relevant
• " = 1
8
10
= 0.2
Exploration Feedback from
the Search Engine
• More diversified results show more mixed results
• Observe the word distribution
• Higher perplexity
µ = 1
total # of the top m frequent non-stop-words 2 Dt
total # of non-stop-words 2 Dt
18
Exploration Feedback from
the Search Engine
19
smoking quitting!q2!
D2! Rank 1: Quit Smoking Hypnosis | Stop Smoking Hypnosis CDs Quit Smoking
Hypnosis Neuro…!
…!
Rank 4: Quit Smoking with Video Hypnosis Home Shopping Cart…!
…!
Rank 10: …!
hypnosis !
• 428 non-stop-words in the top 10 snippets
• the most frequent 5 words:
“smoke”(59),“quit”(34),“hypnosis”(30),“stop”(19),“button”(7)
• µ = 1
59 + 34 + 30 + 19 + 7
428
= 0.36
Put into a POSG Framework
• Partially Observable Stochastic Games (POSGs)
• multiple-agent version of POMDP
• A tuple <S,G,T,R> for States, Agents, Transitions,
Rewards
• G is a tuple too, for a set of agents , each is
<A,O,B>
• Actions, Observations, and Beliefs
20
Observation-Action Pairs
• indicates at time t that we can observe how the
user has browsed the previously retrieved search
results, clicked the documents, and reformulated
the query at the current search iteration.
• indicates that, at time t, the search engine
selects among its search algorithm options,
executes the search algorithms, and provides a
ranked list of search results.
21
(ot, at)
ot
at
Expectation Maximization
(EM) to Learn the Model
• Starts with a random policy
• At the Expectation step
• Compute the decision-making distribution
• Index the most likely decision by j
• A new policy is estimated by finding the best policy at step t
given the current estimates of model parameters and
• At the Maximization step
• Re-compute model parameters based on new estimate of the
policy
22
23
Experiments
• TREC 2012, 2013, 2014 Session Track data
• Immediate Search Effectiveness
• nDCG@10 at each search iteration
• TREC used nDCG@10 at the last search
interaction
24
Baselines
• Lemur: Lemur worked on the last query in a session
• Lemur+all: Lemur concatenating all the queries in a session
• QCM: query change model
• Win-win short: Win-Win uses short-term feedback, e.g. user
clicks, as rewards
• Win-win long: Win-Win uses long-term feedback, nDCG, as
rewards
• served as a performance upper bound
25
TREC 2012 Session
26
• fl performs the
best besides
winwin-long
• lemur+all, qcm,
winwin-long and
fl monotonically
increase over
iterations
• winwin-long > fl,
qcm, lemur+all >
winwin-short
>lemur > original
TREC 2013 Session
27
• Performance
boost at around
2nd iteration and
converge at the
5~6th iterations
• First a few queries
are more
representative
TREC 2014 Session
28
• fl achieves
significant
nDCG@10
improvement over
qcm on TREC’13
and TREC’14
–A new thinking
“The search engine and the user are
equal partners.”
29
Based on that, this paper
• Models the two-way communication between the
two partners on
• relevance
• desire to explore
• Proposes an EM algorithm for learning the best
policy in this framework
30
Look into the future
• Reinforcement-learning-style methods are good for
modeling information seeking
• A lot of room to study the user and the search engine
interaction in a generative way
• The thinking of equal partnership and two-way
communication could be able to generate a set of new
methods and algorithms
• on not only retrieval, but other related fields
• Exciting!!
31
Thank You!
• Email: huiyang@cs.georgetown.edu
• Group Page: Infosense at http://
infosense.cs.georgetown.edu/
• Dynamic IR Website: http://www.dynamic-ir-modeling.org/
• Live Online Search Engine: http://dumplingproject.org
• Upcoming Book: Dynamic Information Retrieval Modeling
• TREC 2015 Dynamic Domain Track: http://trec-dd.org/
32

More Related Content

Similar to Learning to Reinforce Search Effectiveness

Hijacking Treejack UXPA 2023 Talk.pptx.pdf
Hijacking Treejack UXPA 2023 Talk.pptx.pdfHijacking Treejack UXPA 2023 Talk.pptx.pdf
Hijacking Treejack UXPA 2023 Talk.pptx.pdfshalgren
 
Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...
Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...
Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...Jennifer Romano Bergstrom
 
A Practical Guide to Actionable Audience Research
A Practical Guide to Actionable Audience ResearchA Practical Guide to Actionable Audience Research
A Practical Guide to Actionable Audience ResearchDevon Smith
 
An Intro to Jobs-To-Be-Done
An Intro to Jobs-To-Be-DoneAn Intro to Jobs-To-Be-Done
An Intro to Jobs-To-Be-DoneAndrei Radulescu
 
The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...
The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...
The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...Codemotion
 
Surveys that work: using questionnaires to gather useful data, November 2010
Surveys that work: using questionnaires to gather useful data, November 2010Surveys that work: using questionnaires to gather useful data, November 2010
Surveys that work: using questionnaires to gather useful data, November 2010Caroline Jarrett
 
URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2
URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2
URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2URBACT
 
Intro to Agile and Lean UX
Intro to Agile and Lean UXIntro to Agile and Lean UX
Intro to Agile and Lean UXJacklyn Burgan
 
URBACT Summer University 2013 - Labs - Attractive cities for young people - S...
URBACT Summer University 2013 - Labs - Attractive cities for young people - S...URBACT Summer University 2013 - Labs - Attractive cities for young people - S...
URBACT Summer University 2013 - Labs - Attractive cities for young people - S...URBACT
 
Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0
Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0
Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0NTEN
 
Getting started with a systematic review: developing your review question.
Getting started with a systematic review: developing your review question.Getting started with a systematic review: developing your review question.
Getting started with a systematic review: developing your review question.University of Liverpool Library
 
LKCE18 Dimitar Bakardziev - Kanban Policy Game
LKCE18 Dimitar Bakardziev - Kanban Policy GameLKCE18 Dimitar Bakardziev - Kanban Policy Game
LKCE18 Dimitar Bakardziev - Kanban Policy GameLean Kanban Central Europe
 
ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...
ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...
ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...Kazutoshi Umemoto
 
Ch. 7 finish and review
Ch. 7 finish and reviewCh. 7 finish and review
Ch. 7 finish and reviewjbnx
 

Similar to Learning to Reinforce Search Effectiveness (20)

Hijacking Treejack UXPA 2023 Talk.pptx.pdf
Hijacking Treejack UXPA 2023 Talk.pptx.pdfHijacking Treejack UXPA 2023 Talk.pptx.pdf
Hijacking Treejack UXPA 2023 Talk.pptx.pdf
 
Unit1 ed572seminar
Unit1 ed572seminarUnit1 ed572seminar
Unit1 ed572seminar
 
Robots, Testing and LAST
Robots, Testing and LASTRobots, Testing and LAST
Robots, Testing and LAST
 
Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...
Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...
Beyond Eye Tracking: Using User Temperature, Rating Dials, and Facial Analysi...
 
A Practical Guide to Actionable Audience Research
A Practical Guide to Actionable Audience ResearchA Practical Guide to Actionable Audience Research
A Practical Guide to Actionable Audience Research
 
An Intro to Jobs-To-Be-Done
An Intro to Jobs-To-Be-DoneAn Intro to Jobs-To-Be-Done
An Intro to Jobs-To-Be-Done
 
Chapter 7.pdf
Chapter 7.pdfChapter 7.pdf
Chapter 7.pdf
 
qury.pdf
qury.pdfqury.pdf
qury.pdf
 
The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...
The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...
The Secrets of High Performance: Science Edition - Nicole Forsgren - Codemoti...
 
Surveys that work: using questionnaires to gather useful data, November 2010
Surveys that work: using questionnaires to gather useful data, November 2010Surveys that work: using questionnaires to gather useful data, November 2010
Surveys that work: using questionnaires to gather useful data, November 2010
 
URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2
URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2
URBACT Summer University 2013 - Labs - 21st Century Public Services - Session 2
 
Search quality in practice
Search quality in practiceSearch quality in practice
Search quality in practice
 
Intro to Agile and Lean UX
Intro to Agile and Lean UXIntro to Agile and Lean UX
Intro to Agile and Lean UX
 
URBACT Summer University 2013 - Labs - Attractive cities for young people - S...
URBACT Summer University 2013 - Labs - Attractive cities for young people - S...URBACT Summer University 2013 - Labs - Attractive cities for young people - S...
URBACT Summer University 2013 - Labs - Attractive cities for young people - S...
 
Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0
Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0
Nerd, Geek, and Gear Herding: Technical Management Techniques for Managers v 2.0
 
Getting started with a systematic review: developing your review question.
Getting started with a systematic review: developing your review question.Getting started with a systematic review: developing your review question.
Getting started with a systematic review: developing your review question.
 
LKCE18 Dimitar Bakardziev - Kanban Policy Game
LKCE18 Dimitar Bakardziev - Kanban Policy GameLKCE18 Dimitar Bakardziev - Kanban Policy Game
LKCE18 Dimitar Bakardziev - Kanban Policy Game
 
ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...
ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...
ScentBar: A Query Suggestion Interface Visualizing the Amount of Missed Relev...
 
Thinkin overview of possibilities
Thinkin overview of possibilitiesThinkin overview of possibilities
Thinkin overview of possibilities
 
Ch. 7 finish and review
Ch. 7 finish and reviewCh. 7 finish and review
Ch. 7 finish and review
 

Recently uploaded

Biopesticide (2).pptx .This slides helps to know the different types of biop...
Biopesticide (2).pptx  .This slides helps to know the different types of biop...Biopesticide (2).pptx  .This slides helps to know the different types of biop...
Biopesticide (2).pptx .This slides helps to know the different types of biop...RohitNehra6
 
G9 Science Q4- Week 1-2 Projectile Motion.ppt
G9 Science Q4- Week 1-2 Projectile Motion.pptG9 Science Q4- Week 1-2 Projectile Motion.ppt
G9 Science Q4- Week 1-2 Projectile Motion.pptMAESTRELLAMesa2
 
Unlocking the Potential: Deep dive into ocean of Ceramic Magnets.pptx
Unlocking  the Potential: Deep dive into ocean of Ceramic Magnets.pptxUnlocking  the Potential: Deep dive into ocean of Ceramic Magnets.pptx
Unlocking the Potential: Deep dive into ocean of Ceramic Magnets.pptxanandsmhk
 
Stunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCR
Stunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCRStunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCR
Stunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCRDelhi Call girls
 
Presentation Vikram Lander by Vedansh Gupta.pptx
Presentation Vikram Lander by Vedansh Gupta.pptxPresentation Vikram Lander by Vedansh Gupta.pptx
Presentation Vikram Lander by Vedansh Gupta.pptxgindu3009
 
Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?Patrick Diehl
 
Broad bean, Lima Bean, Jack bean, Ullucus.pptx
Broad bean, Lima Bean, Jack bean, Ullucus.pptxBroad bean, Lima Bean, Jack bean, Ullucus.pptx
Broad bean, Lima Bean, Jack bean, Ullucus.pptxjana861314
 
GFP in rDNA Technology (Biotechnology).pptx
GFP in rDNA Technology (Biotechnology).pptxGFP in rDNA Technology (Biotechnology).pptx
GFP in rDNA Technology (Biotechnology).pptxAleenaTreesaSaji
 
Botany 4th semester series (krishna).pdf
Botany 4th semester series (krishna).pdfBotany 4th semester series (krishna).pdf
Botany 4th semester series (krishna).pdfSumit Kumar yadav
 
Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...
Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...
Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...jana861314
 
Nightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43b
Nightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43bNightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43b
Nightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43bSérgio Sacani
 
Zoology 4th semester series (krishna).pdf
Zoology 4th semester series (krishna).pdfZoology 4th semester series (krishna).pdf
Zoology 4th semester series (krishna).pdfSumit Kumar yadav
 
Botany krishna series 2nd semester Only Mcq type questions
Botany krishna series 2nd semester Only Mcq type questionsBotany krishna series 2nd semester Only Mcq type questions
Botany krishna series 2nd semester Only Mcq type questionsSumit Kumar yadav
 
Grafana in space: Monitoring Japan's SLIM moon lander in real time
Grafana in space: Monitoring Japan's SLIM moon lander  in real timeGrafana in space: Monitoring Japan's SLIM moon lander  in real time
Grafana in space: Monitoring Japan's SLIM moon lander in real timeSatoshi NAKAHIRA
 
Boyles law module in the grade 10 science
Boyles law module in the grade 10 scienceBoyles law module in the grade 10 science
Boyles law module in the grade 10 sciencefloriejanemacaya1
 
Nanoparticles synthesis and characterization​ ​
Nanoparticles synthesis and characterization​  ​Nanoparticles synthesis and characterization​  ​
Nanoparticles synthesis and characterization​ ​kaibalyasahoo82800
 
Isotopic evidence of long-lived volcanism on Io
Isotopic evidence of long-lived volcanism on IoIsotopic evidence of long-lived volcanism on Io
Isotopic evidence of long-lived volcanism on IoSérgio Sacani
 
Chemistry 4th semester series (krishna).pdf
Chemistry 4th semester series (krishna).pdfChemistry 4th semester series (krishna).pdf
Chemistry 4th semester series (krishna).pdfSumit Kumar yadav
 
Formation of low mass protostars and their circumstellar disks
Formation of low mass protostars and their circumstellar disksFormation of low mass protostars and their circumstellar disks
Formation of low mass protostars and their circumstellar disksSérgio Sacani
 

Recently uploaded (20)

Biopesticide (2).pptx .This slides helps to know the different types of biop...
Biopesticide (2).pptx  .This slides helps to know the different types of biop...Biopesticide (2).pptx  .This slides helps to know the different types of biop...
Biopesticide (2).pptx .This slides helps to know the different types of biop...
 
G9 Science Q4- Week 1-2 Projectile Motion.ppt
G9 Science Q4- Week 1-2 Projectile Motion.pptG9 Science Q4- Week 1-2 Projectile Motion.ppt
G9 Science Q4- Week 1-2 Projectile Motion.ppt
 
Unlocking the Potential: Deep dive into ocean of Ceramic Magnets.pptx
Unlocking  the Potential: Deep dive into ocean of Ceramic Magnets.pptxUnlocking  the Potential: Deep dive into ocean of Ceramic Magnets.pptx
Unlocking the Potential: Deep dive into ocean of Ceramic Magnets.pptx
 
Stunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCR
Stunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCRStunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCR
Stunning ➥8448380779▻ Call Girls In Panchshil Enclave Delhi NCR
 
Presentation Vikram Lander by Vedansh Gupta.pptx
Presentation Vikram Lander by Vedansh Gupta.pptxPresentation Vikram Lander by Vedansh Gupta.pptx
Presentation Vikram Lander by Vedansh Gupta.pptx
 
Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?
 
Broad bean, Lima Bean, Jack bean, Ullucus.pptx
Broad bean, Lima Bean, Jack bean, Ullucus.pptxBroad bean, Lima Bean, Jack bean, Ullucus.pptx
Broad bean, Lima Bean, Jack bean, Ullucus.pptx
 
CELL -Structural and Functional unit of life.pdf
CELL -Structural and Functional unit of life.pdfCELL -Structural and Functional unit of life.pdf
CELL -Structural and Functional unit of life.pdf
 
GFP in rDNA Technology (Biotechnology).pptx
GFP in rDNA Technology (Biotechnology).pptxGFP in rDNA Technology (Biotechnology).pptx
GFP in rDNA Technology (Biotechnology).pptx
 
Botany 4th semester series (krishna).pdf
Botany 4th semester series (krishna).pdfBotany 4th semester series (krishna).pdf
Botany 4th semester series (krishna).pdf
 
Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...
Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...
Traditional Agroforestry System in India- Shifting Cultivation, Taungya, Home...
 
Nightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43b
Nightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43bNightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43b
Nightside clouds and disequilibrium chemistry on the hot Jupiter WASP-43b
 
Zoology 4th semester series (krishna).pdf
Zoology 4th semester series (krishna).pdfZoology 4th semester series (krishna).pdf
Zoology 4th semester series (krishna).pdf
 
Botany krishna series 2nd semester Only Mcq type questions
Botany krishna series 2nd semester Only Mcq type questionsBotany krishna series 2nd semester Only Mcq type questions
Botany krishna series 2nd semester Only Mcq type questions
 
Grafana in space: Monitoring Japan's SLIM moon lander in real time
Grafana in space: Monitoring Japan's SLIM moon lander  in real timeGrafana in space: Monitoring Japan's SLIM moon lander  in real time
Grafana in space: Monitoring Japan's SLIM moon lander in real time
 
Boyles law module in the grade 10 science
Boyles law module in the grade 10 scienceBoyles law module in the grade 10 science
Boyles law module in the grade 10 science
 
Nanoparticles synthesis and characterization​ ​
Nanoparticles synthesis and characterization​  ​Nanoparticles synthesis and characterization​  ​
Nanoparticles synthesis and characterization​ ​
 
Isotopic evidence of long-lived volcanism on Io
Isotopic evidence of long-lived volcanism on IoIsotopic evidence of long-lived volcanism on Io
Isotopic evidence of long-lived volcanism on Io
 
Chemistry 4th semester series (krishna).pdf
Chemistry 4th semester series (krishna).pdfChemistry 4th semester series (krishna).pdf
Chemistry 4th semester series (krishna).pdf
 
Formation of low mass protostars and their circumstellar disks
Formation of low mass protostars and their circumstellar disksFormation of low mass protostars and their circumstellar disks
Formation of low mass protostars and their circumstellar disks
 

Learning to Reinforce Search Effectiveness

  • 1. Learning to Reinforce Search Effectiveness Jiyun Luo, Xuchu Dong, Grace Hui Yang Georgetown University ICTIR 2015 1
  • 3. But you are not alone: Search with a Partner • A teamwork • Share a common goal • find relevant documents • satisfy long term goal • Equal partners • not just being an assistant to the user • but also providing influence • Cooperative exploration 3
  • 4. Key Idea: Cooperative Exporation • The two parties • They talk and they listen • keep exchanging their ideas • Take turns to lead the search to the direction each of them would like this collaboration to go • Also considering the other’s opinion 4
  • 5. t=1 Query q1=“hydropower efficiency” Messages:Messages: See my new query. Let’s explore! 5 Example is from TREC 2014 Session 52
  • 6. t=1 Query q1=“hydropower efficiency” Retrieved docs D1 “…renewable energy…” Messages:Messages: Check it out! Documents I’ve ranked high are relevant See my new query. Let’s explore! 6 Example is from TREC 2014 Session 52
  • 7. t=1 t=2 Query q1=“hydropower efficiency” Clicked d2 in D1 Query q2=“hydropower environment” Retrieved docs D1 “…renewable energy…” Messages:Messages: Check it out! Documents I’ve ranked high are relevant See my new query. Let’s explore! Documents I’ve clicked look relevant! My new query is on another subtopic. Let’s explore 7 Example is from TREC 2014 Session 52
  • 8. t=1 t=2 Retrieved docs D2 Query q1=“hydropower efficiency” Clicked d2 in D1 Query q2=“hydropower environment” Retrieved docs D1 “…renewable energy…” Messages:Messages: Check it out! Documents I’ve ranked high are relevant Check it out! Documents I’ve ranked high are relevant. See my new query. Let’s explore! Documents I’ve clicked look relevant! My new query is on another subtopic. Let’s explore 8 Example is from TREC 2014 Session 52
  • 9. t=1 t=2 t=3 Retrieved docs D2 Query q1=“hydropower efficiency” Clicked d2 in D1 Query q2=“hydropower environment” Clicked d2 in D2 Query q3=“hydropower damage” Retrieved docs D1 “…renewable energy…” Messages:Messages: Check it out! Documents I’ve ranked high are relevant Check it out! Documents I’ve ranked high are relevant. See my new query. Let’s explore! Documents I’ve clicked look relevant! Documents I’ve clicked look relevant! My new query is on another subtopic. Let’s explore My new query is still on the same subtopic. Let’s find out more about it. 9 Example is from TREC 2014 Session 52
  • 10. t=1 t=2 t=3 … t=4 Retrieved docs D2 Retrieved docs D3 Query q1=“hydropower efficiency” Clicked d2 in D1 Query q2=“hydropower environment” Clicked d2 in D2 Query q3=“hydropower damage” Retrieved docs D1 “…renewable energy…” Messages:Messages: Check it out! Documents I’ve ranked high are relevant Check it out! Documents I’ve ranked high are relevant. See my new query. Let’s explore! Documents I’ve clicked look relevant! Documents I’ve clicked look relevant! Want to explore? I’ve diversified my results. My new query is on another subtopic. Let’s explore My new query is still on the same subtopic. Let’s find out more about it. 10 Example is from TREC 2014 Session 52
  • 11. Opinions about Two Things • Relevance • Which documents (that you have just marked/ retrieved/recommended) are relevant • Desire of Exploration • How exploratory I want us to be, as a team 11
  • 12. How to Express the Opinions/ Feedback • Relevance is “demonstrated by examples”: • Query is a piece of short text sent from the user • Clicked snippets/documents are long pieces of text sent from user • Documents are long text sent from the search engine • Desire of Exploration is shown by • Query changes • Diversified results 12
  • 13. A Contextual Bandit Formulation of a Decision-Making Distribution P(relevant) = 1 ✏ P(explore) = µ P(J = RE|o, a, ⇡⇤ ) = (1 ✏)µ P(J = NRE|o, a, ⇡⇤ ) = ✏µ P(J |o, a, ⇡⇤ ) = P(relevant)P(explore) P(J = RNE|o, a, ⇡⇤ ) = (1 ✏)(1 µ) P(J = NRNE|o, a, ⇡⇤ ) = ✏(1 µ) 13
  • 14. Relevance Feedback from the User • 1 SAT-Clicked out of 10 retrieved, ✏ = 1 # of SAT-Clicked documents 2 Dt 1 # of returned documents 2 Dt 1 14 " = 1 1 10 = 0.9 smoking quitting!q2! hypnosis ! Rank 1: Easy Ways to Quit Smoking | Quit Smoking Help … ! …! Rank 3: Quit Smoking Toolbox - Quit Smoking - Nicotine Addiction …! …! Rank 6: Quit Smoking Hypnosis, Stop Smoking Hypnosis CDs…! …! Rank 10: …! SAT-Clicked. ! Dwell time: 40 seconds! D1!
  • 15. Exploration Feedback from the User • 1 query change , 3 terms in the new query in total µ = 1 # of query changes 2 Dt 1 # of permutations of query terms 2 Dt 1 15 smoking quitting!q2! hypnosis ! +∆q" Rank 1: Easy Ways to Quit Smoking | Quit Smoking Help … ! …! Rank 3: Quit Smoking Toolbox - Quit Smoking - Nicotine Addiction …! …! Rank 6: Quit Smoking Hypnosis, Stop Smoking Hypnosis CDs…! …! Rank 10: …! D1! Query reformulation using words in previous search results! 2 Dt 1 µ = 1 1 3! = 0.83
  • 16. Relevance Feedback from the Search Engine • Highly scored documents • Needs consistency in ranking scores • Could be hard to get • Highly ranked documents ✏ = 1 # of relevant documents 2 top n retrieved n 16
  • 17. Relevance Feedback from the Search Engine 17 smoking quitting!q2! D2! Rank 1: Quit Smoking Hypnosis | Stop Smoking Hypnosis CDs Quit Smoking Hypnosis Neuro…! …! Rank 4: Quit Smoking with Video Hypnosis Home Shopping Cart…! …! Rank 10: …! hypnosis ! • 8 out of 10 top retrieved documents are relevant • " = 1 8 10 = 0.2
  • 18. Exploration Feedback from the Search Engine • More diversified results show more mixed results • Observe the word distribution • Higher perplexity µ = 1 total # of the top m frequent non-stop-words 2 Dt total # of non-stop-words 2 Dt 18
  • 19. Exploration Feedback from the Search Engine 19 smoking quitting!q2! D2! Rank 1: Quit Smoking Hypnosis | Stop Smoking Hypnosis CDs Quit Smoking Hypnosis Neuro…! …! Rank 4: Quit Smoking with Video Hypnosis Home Shopping Cart…! …! Rank 10: …! hypnosis ! • 428 non-stop-words in the top 10 snippets • the most frequent 5 words: “smoke”(59),“quit”(34),“hypnosis”(30),“stop”(19),“button”(7) • µ = 1 59 + 34 + 30 + 19 + 7 428 = 0.36
  • 20. Put into a POSG Framework • Partially Observable Stochastic Games (POSGs) • multiple-agent version of POMDP • A tuple <S,G,T,R> for States, Agents, Transitions, Rewards • G is a tuple too, for a set of agents , each is <A,O,B> • Actions, Observations, and Beliefs 20
  • 21. Observation-Action Pairs • indicates at time t that we can observe how the user has browsed the previously retrieved search results, clicked the documents, and reformulated the query at the current search iteration. • indicates that, at time t, the search engine selects among its search algorithm options, executes the search algorithms, and provides a ranked list of search results. 21 (ot, at) ot at
  • 22. Expectation Maximization (EM) to Learn the Model • Starts with a random policy • At the Expectation step • Compute the decision-making distribution • Index the most likely decision by j • A new policy is estimated by finding the best policy at step t given the current estimates of model parameters and • At the Maximization step • Re-compute model parameters based on new estimate of the policy 22
  • 23. 23
  • 24. Experiments • TREC 2012, 2013, 2014 Session Track data • Immediate Search Effectiveness • nDCG@10 at each search iteration • TREC used nDCG@10 at the last search interaction 24
  • 25. Baselines • Lemur: Lemur worked on the last query in a session • Lemur+all: Lemur concatenating all the queries in a session • QCM: query change model • Win-win short: Win-Win uses short-term feedback, e.g. user clicks, as rewards • Win-win long: Win-Win uses long-term feedback, nDCG, as rewards • served as a performance upper bound 25
  • 26. TREC 2012 Session 26 • fl performs the best besides winwin-long • lemur+all, qcm, winwin-long and fl monotonically increase over iterations • winwin-long > fl, qcm, lemur+all > winwin-short >lemur > original
  • 27. TREC 2013 Session 27 • Performance boost at around 2nd iteration and converge at the 5~6th iterations • First a few queries are more representative
  • 28. TREC 2014 Session 28 • fl achieves significant nDCG@10 improvement over qcm on TREC’13 and TREC’14
  • 29. –A new thinking “The search engine and the user are equal partners.” 29
  • 30. Based on that, this paper • Models the two-way communication between the two partners on • relevance • desire to explore • Proposes an EM algorithm for learning the best policy in this framework 30
  • 31. Look into the future • Reinforcement-learning-style methods are good for modeling information seeking • A lot of room to study the user and the search engine interaction in a generative way • The thinking of equal partnership and two-way communication could be able to generate a set of new methods and algorithms • on not only retrieval, but other related fields • Exciting!! 31
  • 32. Thank You! • Email: huiyang@cs.georgetown.edu • Group Page: Infosense at http:// infosense.cs.georgetown.edu/ • Dynamic IR Website: http://www.dynamic-ir-modeling.org/ • Live Online Search Engine: http://dumplingproject.org • Upcoming Book: Dynamic Information Retrieval Modeling • TREC 2015 Dynamic Domain Track: http://trec-dd.org/ 32