Challenges in online learning to rank for information retrieval thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Challenges in online learning to rank for information retrieval

Published on Nov 07, 20132671 Views

Online learning to rank for information retrieval (IR) aims to enable search systems to learn directly from interactions with their users. In our recent work, we explore formulations based on reinforc

Related categories

Chapter list

Challenges in Online Learning to Rank for Information Retrieval (IR)00:00
Labels for Large-Scale Learning00:19
Example: Web Search00:41
Manual Annotations?02:03
Labels for Large-Scale Learning - 102:48
Labels for Large-Scale Learning - 203:31
Learning from User Interactions:Challenges03:42
Interpreting User Interactions04:29
Position Bias04:35
Absolute Metrics05:49
Interpreting clicks as pairwise feedback07:05
Interpreting clicks as listwise feedback07:55
Summary: Interpreting interactions as feedback for learning09:50
Balancing exploration and exploitation10:11
Problem Formulation10:24
The Online Learning to Rank Challenge11:26
Question12:19
Pairwise Learning to Rank for IR13:00
Balancing Exploration and Exploitation in Pairwise Learning13:39
Listwise Learning to Rank for IR14:38
Balancing Exploration and Exploitation in Listwise Learning15:37
2 Approaches –Summary16:35
Experiments17:14
Results: Online Performance Pairwise Approach18:35
Results: Offline Performance Pairwise Approach20:05
Results: Online Performance Listwise Approach21:44
Results: Offline Performance Listwise Approach22:52
Summary23:44
Related / Ongoing Work24:32
Outlook: Smart Exploration25:37