Tomáš Tunys
Jan Šedivý�
LEARNING TO RANK
FROM
IMPLICIT USER FEEDBACK
dissertation thesis proposal
Presentation Outline
Introduction and Motivation
Needles in the Haystack
“With No Signs of Slowing, the Data Keeps Growing”
2.1 BILLION PEOPLE
source: http://www.domo.com/learn/
Needles in the Haystack
“With No Signs of Slowing, the Data Keeps Growing”
2.4 BILLION PEOPLE
source: http://www.domo.com/learn/
Corollary: Satisfying Information Needs Becomes Harder
Corollary: Satisfying Information Needs Becomes Harder
What is Learning To Rank?
Learning To Rank
Searching for relevant documents
Documents
Document d1
Document d2
Document d3
Document d4
Document d5
Feature Vectors
(... q + d1 …)
(... q + d2 …)
(... q + d3 …)
(... q + d4 …)
(... q + d5 …)
Query q
Relevance Labels
relevance y1
relevance y2
relevance y3
relevance y4
relevance y5
Learning To Rank
Searching for relevant documents
Documents
Document d1
Document d2
Document d3
Document d4
Document d5
Feature Vectors
(... q + d1 …)
(... q + d2 …)
(... q + d3 …)
(... q + d4 …)
(... q + d5 …)
Query q
Relevance Labels
relevance y1
relevance y2
relevance y3
relevance y4
relevance y5
?
?
Learning To Rank
Searching for relevant documents
Documents
Document d1
Document d2
Document d3
Document d4
Document d5
Feature Vectors
(... q + d1 …)
(... q + d2 …)
(... q + d3 …)
(... q + d4 …)
(... q + d5 …)
Query q
Relevance Labels
relevance y1
relevance y2
relevance y3
relevance y4
relevance y5
Ranking Model
Learning phase
Learning To Rank
Searching for relevant documents
Documents
Document d1
Document d2
Document d3
Document d4
Document d5
Feature Vectors
(... q + d1 …)
(... q + d2 …)
(... q + d3 …)
(... q + d4 …)
(... q + d5 …)
Query q
Ranking Model
(inference)
score s1
score s2
score s3
score s4
score s5
Production phase
Learning to Rank
Formulation, Evaluation, Algorithms
Learning To Rank: Problem Formulation
Empirical Risk Minimization
Q
performance measure
Evaluation Measures
Normalized Discounted Cumulative Gain
Evaluation Measures
Training Data Problem
Very expensive, time consuming, and of disputable quality.
Learning To Rank
Types of Algorithms
Learning To Rank
Types of Algorithms
Learning To Rank From Implicit Feedback
Learning To Rank
Exploiting User Implicit Feedback
Learning To Rank
Exploiting User Implicit Feedback
Learning To Rank
Exploiting User Implicit Feedback
Current and Future Work
Current and Future Work
Current and Future Work