The Dynamics of Textual Content on Social Media Restricted; Files Only

Zhong, Ning (Spring 2019)

Permanent URL: https://etd.library.emory.edu/concern/etds/qv33rx69g?locale=en
Published

Abstract

While social media have emerged as open sources of insights for both marketing

researchers and practitioners, much of the work on the dynamics in social media activity has

focused on numeric metrics such as volume and valence. Exiting literature on the usergenerated

content (UGC) on social media has begun to explore its potential to yield

marketing insights, but little has been done to consider how the textual content on social

media may shift over time. The goal of this dissertation work is to find out how the textbased

UGC on social media evolves over time by extending the topic modeling framework of

latent Dirichlet allocation (LDA) in three empirical scenarios. In the first essay, a discretestate

dynamic topic model that incorporates multiple latent changepoints is developed to

capture the underlying shifts of textual content that relates to a brand on social media

around an event, such as a brand crisis, a new product release, or breaking news. This

model may be used by marketers to actively monitor online conversations surrounding their

own brands by detecting changes in the topics discussed on social media. In the second

essay, a continuous-state dynamic topic model is proposed to examine the evolution of topic

prevalence and evaluations in customer reviews on a multi-generational product. The

findings show that the concerns of the review contributors at the early stage of product

lifecycle are different from those of the review contributors at the later stage.

Table of Contents

INTRODUCTION......................................................................................................................... 1

ESSAY 1

Capturing Changes in Social Media Content: A Multiple Latent Changepoint Topic Model ........ 6

Abstract ....................................................................................................................................... 6

Introduction ................................................................................................................................. 7

Related Literature ....................................................................................................................... 9

Model Development ................................................................................................................. 13

Empirical Applications ............................................................................................................. 20

Discussion ................................................................................................................................. 32

References ................................................................................................................................. 36

Appendix A.1. MCMC Algorithm for LDA-LC ....................................................................... 55

Appendix A.2. Simulation Study .............................................................................................. 59

Appendix A.3. Most Relevant Words and Prevalence of Topics .............................................. 62

ESSAY 2

The Evolution of Online Reviews: A Dynamic Topic Model for Multiple Text Streams ............. 66

Abstract ..................................................................................................................................... 66

Introduction ............................................................................................................................... 67

Related Literature ..................................................................................................................... 70

Data ........................................................................................................................................... 72

Model Development ................................................................................................................. 75

Results ....................................................................................................................................... 81

Conclusion ................................................................................................................................ 88

Reference .................................................................................................................................. 92

Appendix A.1. Collapsed Gibbs Sampler for SLDA .............................................................. 106

Appendix A.2. MCMC Algorithm for SLDA-MS .................................................................. 108

Appendix A.3. In-sample Model Fit ........................................................................................112

Appendix A.4. Results and Discussions of the Remaining Categories ...................................113

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
Department
Subfield / Discipline
Degree
Submission
Language
  • English
Research field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Last modified No preview

Primary PDF

Supplemental Files