Modeling User Attention and Interaction on the Web Public

Abstract

Analysis of user attention and Web page examination behavior, collected with specialized eye tracking equipment, has offered numerous insights about how users examine content online. Unfortunately, eye tracking technology is currently available for relatively small scale user studies, due to its high costs and the effort associated with participant recruitment. This thesis develops several alternatives to eye tracking for studying user attention and behavior. We start by introducing ViewSer - a method based on idea of restricted focus viewing that allows measuring attention for thousands of participants. Then, we develop a probabilistic model that infers most likely position of user's gaze on the screen from user interactions and Web page content. Our model outperforms current state of the art for gaze position prediction that only uses behavioral signals or information about Web page visual content. In addition to the methodological contributions, this thesis develops several important applications in Web search and medical domain. First, we describe a scalable approach for extracting frequent mouse cursor movement patterns from large scale cursor data. These patterns could be used to improve quality of search result relevance estimation and search result ranking. Second, we show that attention measured with cursor and viewport position could be used to improve automatic Web page summarization. Lastly, we demonstrate an important medical application of restricted focus viewing, for automated diagnostic of memory impairment that could be administered remotely over the Internet anywhere in the world. Together, the techniques developed and evaluated in this thesis substantially advance the state of the art of user attention modeling and enable novel practical applications.

Table of Contents

1 Introduction

1.1 Background and Motivation 1

1.2 Contributions 5

2 Related Work

2.1 Eye Movement Data and User Interfaces 7

2.2 Eye Movements and Information Processing 8

2.3 Modeling Attention from Cursor Interactions 9

2.4 User Interactions in Web Search 10

3 Restricted Focus Viewing

3.1 Background and Motivation 12

3.2 ViewSer Implementation for Web Pages 14

3.3 Validating Viewser for Search Result Pages 16

3.4 Summary 22

4 Discovery of Cursor Movement and Interaction Patterns

4.1 Background and Motivation 23

4.2 Problem Statement 25

4.3 Cursor Motif Discovery 26

4.3.1 Candidate generation and pre-processing 26

4.3.2 Distance Measure 26

4.3.3 Candidate Similarity Computation 27

4.3.4 Scaling Up Motif Discovery 28

4.4 Scalable Motif Discovery 30

4.4.1 Experimental Setup and Dataset 30

4.4.2 Evaluating Distance Measure Learning 31

4.4.3 Runtime Performance 32

4.5 Discovered Cursor Motifs 32

4.6 Summary 34

5 Modeling Attention from Page Content and Interactions

5.1 Motivation 35

5.2 Attention Tracking from Cursor Movements 35

5.2.1 Notation and Data 36

5.2.2 Linear Regression 37

5.2.3 Non-Linear Regression 38

5.3 Effect of Web Page Content on Attention Distribution 38

5.4 Computational Visual Saliency 40

5.4.1 Graph Based Visual Saliency 42

5.5 MICS: Mixture of Interactions and Content Saliency 42

5.5.1 Definition 43

5.5.2 Training 44

5.5.3 Inference 45

5.6 Experiments 46

5.6.1 Model Implementation 46

5.6.2 Extracting Prominent Web Page Elements 46

5.6.3 Content and Interaction Features 48

5.6.4 Baseline Interaction Features 48

5.6.5 Evaluation Metrics 50

5.6.6 Comparison of MICS and GBVS 52

5.6.7 Comparison of MICS and Regression Models 53

5.7 Summary 54

6 Applications to Web Search

6.1 Restricted Focus Viewing 55

6.1.1 Relevance ratings collection 55

6.1.2 Snippet Attractiveness 56

6.1.3 Search Result Re-ranking 57

6.1.4 Detecting Bad Snippets 58

6.2 Motifs for Relevance Prediction and Ranking 61

6.3 Attention Biased Document Summarization 65

6.3.1 Problem Statement 65

6.3.2 Approach 66

6.3.3 Page Examination Behavior Logging 67

6.3.4 Behavior-Biased Snippet Generation 67

6.3.5 Data Collection and Experimental Setup 71

6.3.6 Results 72

6.4 Summary 77

7 Applications of Attention Tracking in Medical Domain

7.1 Background and Motivation 81

7.1.1 Data 83

7.1.2 Eye Movement Features 83

7.1.3 Classifier Evaluation Procedure 84

7.1.4 Results 85

7.2 Web based Visual Paired Comparison Task 86

7.2.1 Data 87

7.2.2 Results 90

7.3 Summary 96

8 Conclusions

8.1 Summary of Contributions 98

8.2 Current and Future Work 102

About this Dissertation

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Math and Computer Science
Degree	PhD
Submission	Dissertation
Language	English
Research Field	Computer Science Artificial Intelligence Information Science
Mot-clé	user behavior modeling automatic document summarization data mining Web search visual paired comparison online behavioral test user interactions mouse cursor tracking attention modeling eye tracking information retrieval motif extraction
Committee Chair / Thesis Advisor	Agichtein, Eugene, Emory University
Committee Members	Zola, Stuart M, Emory University Smola, Alexander, Carnegie Mellon University Wolff, Phillip, Emory University

Dernière modification

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Modeling User Attention and Interaction on the Web ()	2018-08-28 14:24:18 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions