Improving Usefulness of Community Question Answering Services Towards Better Searcher Satisfaction and Question Recommendation Open Access

Abstract

Community-based Question Answering (CQA) sites, such as Yahoo! Answers and Quora, provide a promising way of finding and sharing information online. This thesis aims to improve the usefulness of CQA services towards better searcher satisfaction and question recommendation, by focusing on three important problems ignored in previous work:

(1) How to improve web searcher satisfaction using CQA services. This thesis proposes methods for a novel task of predicting web searcher satisfaction with existing answers in CQA, enabling better ranking of CQA pages for searchers. When searchers fail in web search, they may alternatively ask questions using CQA services. This thesis analyzes users' transition from searching to asking in terms of query and behavior characteristics, providing insights for predicting the transition.

(2) What contextual factors influence answerer behavior in CQA. This thesis analyzes the answerer behavior in a large scale CQA system, and explores when users tend to answer questions and how they tend to choose the questions to answer. Based on a user study, this thesis further explores how relevant web browsing context affects answerers' perceived ability, effort, and willingness to answer a question. The findings could inform the design of more intelligent question recommendation strategies in CQA systems.

(3) How to deploy question recommendation in real-time CQA systems. This thesis develops a scalable real-time CQA system with a mobile interface, which supports different question recommendation strategies. Based on two live user studies, this thesis further conducts both quantitative analysis of user behavior as well as qualitative analysis of user satisfaction with the system. The developed system and the reported analysis offer insights for designing real-time CQA systems and deploying question recommendation.

In summary, the work on predicting searcher satisfaction and understanding the transition from searching to asking would help improve searcher satisfaction using CQA systems, and the work on understanding answerer behavior and building a real-time CQA system would help improve question recommendation in CQA systems, making CQA services more useful.

Table of Contents

1 Introduction 1

1.1 Motivation 1

1.2 Contributions 8

1.3 Organization 11

2 RelatedWork 12

2.1 Improving Searcher Satisfaction using CQA Services 12

2.1.1 Question and Answer Retrieval in CQA 12

2.1.2 Searcher Satisfaction and Switching Behavior 13

2.1.3 Answer Quality and Asker Satisfaction 15

2.1.4 Query and Question Analysis 16

2.1.5 Improving Search Experience using CQA Data 18

2.2 Question Recommendation and Routing in CQA 19

2.3 Understanding User Behavior in CQA 24

2.4 Building Real-Time CQA Systems 25

2.5 Crowdsourcing and Social Networks 29

3 Predicting Web Searcher Satisfaction with Existing Answers 31

3.1 Problem and Approaches 33

3.2 Experimental Setup 41

3.3 Empirical Evaluation 47

3.4 Summary 54

4 Understanding When Searchers Become Askers 56

4.1 Dataset Preparation 60

4.2 Query and Behavior Analysis 62

4.2.1 Characteristics of Queries leading to Questions 62

4.2.2 Searcher Behavior Before Asking Questions 67

4.3 Queries vs. Questions 73

4.4 Question Analysis 78

4.5 Summary 82

5 Understanding Answerer Behavior for Better Question Recommendation 83

5.1 Modeling Answerer Behavior in CQA 84

5.1.1 Temporal Patterns in Answerer Behavior 85

5.1.2 Understanding How Answerers Choose Questions 92

5.2 Exploring Web Browsing Context for CQA 102

5.2.1 Study Design 104

5.2.2 Results 106

5.3 Summary 110

6 Building A Real-Time CQA System 112

6.1 System Overview 114

6.1.1 Front-end: mobile application 116

6.1.2 Back-end: server system 122

6.2 User studies 126

6.2.1 Statistics and survey responses 128

6.2.2 Question types and answer quality 130

6.2.3 Question recommendation strategies: PULL vs. PUSH 135

6.2.4 Recommendation from the main Page: question ranking 136

6.2.5 Recommendation via notification: user ranking 138

6.2.6 Tag ranking 141

6.2.7 Notification Settings 143

6.3 Discussion and Implications 144

6.4 Summary 147

7 Conclusions and Future Work 149

7.1 Summary of Thesis Work 149

7.2 Limitations and Future Work 157

Bibliography 160

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
Keyword	community question answering searcher satisfaction question recommendation
Committee Chair / Thesis Advisor	Lu, James, Emory University Agichtein, Eugene, Emory University
Committee Members	Eisenstein, Jacob, Georgia Institute of Technology Xiong, Li, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Improving Usefulness of Community Question Answering Services Towards Better Searcher Satisfaction and Question Recommendation ()	2018-08-28 12:02:50 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions