The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Detecting user demographics in twitter to inform health trends in social media
Author:	Markson, Christopher R.
View Online:	njit-etd2017-091 (xvi, 113 pages ~ 2.3 MB pdf)
Department:	Department of Informatics
Degree:	Doctor of Philosophy
Program:	Information Systems
Document Type:	Dissertation
Advisory Committee:	Xu, Songhua (Committee chair) Wu, Yi-Fang Brook (Committee member) Bieber, Michael (Committee member) Phan, Hai Nhat (Committee member) Kaufman, L. C. (Committee member) Iyengar, Reethi Narasimhan (Committee member)
Date:	2017-08
Keywords:	Social media Topic extraction Demographic extraction Text mining Public health Twitter
Availability:	Unrestricted
Abstract:	The widespread and popular use of social media and social networking applications offer a promising opportunity for gaining knowledge and insights regarding population health conditions thanks to the diversity and abundance of online user-generated information (UGHI) relating to healthcare and well-being. However, users on social media and social networking sites often do not supply their complete demographic information, which greatly undermines the value of the aforementioned information for health 2.0 research, e.g., for discerning disparities across population groups in certain health conditions. To recover the missing user demographic information, existing methods observe a limited scope of user behaviors, such as word frequencies exhibited in a user’s messages, leading to sub-optimal results. To address the above limitation and improve the performance of inferring missing user demographic information for health 2.0 research, this work proposes a new algorithmic method for extracting a social media user’s gender by exploring and exploiting a comprehensive set of a user’s behaviors on Twitter, including the user’s conversational topic choices, account profile information, and personal information. In addition, this work explores the usage of synonym expansion for detecting social media users’ ethnicities. To better capture a user’s conversational topic choices using standardized hashtags for consistent comparison, this work additionally introduces a new method that automatically generates standardized hashtags for tweets. Even though Twitter is selected as the experimental platform in this study due to its leading position among today’s social networking sites, the proposed method is in principle generically applicable to other social media sites and applications as long as there is a way to access user-generated content on those platforms. When comparing the multi-perspective learning method with the state-of-the-art approaches for gender classification, a gender classification accuracy is observed of 88.6% for the proposed approach compared with 63.4% performance for bag-of-words and 61.4% for the peer method. Additionally, the topical approach introduced in this work outperforms vocabulary-based approach with a smaller dimensionality at 69.4% accuracy. Furthermore, observable usage patterns of the cancer terms are analyzed across the ethnic groups inferred by the proposed algorithmic approaches. Variations among demographic groups are seen in the frequency of term usage during months known to be labeled as cancer awareness months. This work introduces methods that have the potential to serve as a very powerful and important tool in disseminating critical prevention, screening, and treatment messages to the community in real time. Study findings highlight the potential benefits of social media as a tool for detecting demographic differences in cancer-related discussions on social media.