Feifei Zhang | Syracuse University (original) (raw)
Feifei Zhang's research interest is to investigate human beings’ use of information and communication technologies (ICTs), and their social, political, cultural and psychological influence in facilitating and shaping people’s behaviors, especially in interpersonal and political contexts. Her first goal is to test whether new theories can be developed or existing relevant theories can evolve to explain new phenomena in the digitally mediated communication environment. Her second goal is to explore how theories can help design new technologies and features to improve social interaction, enable collaboration, enhance individuals' life satisfaction, and facilitate political participation and civic engagement.
Feifei Zhang uses mixed research methods to collect, mine and analyze both online and offline data, primarily quantitative research methods and computational approaches, including content analysis, survey, text mining, natural language process and machine learning.
Supervisors: Dr. Jennifer Stromer-Galley, Dr. Bei Yu, Dr. Bryan Semaan, and Dr. Qiu Wang
less
Uploads
Papers by Feifei Zhang
Proceedings of the 8th International Conference on Social Media & Society, 2017
In this paper, we introduce a lexicon-based method for identifying political topics in social med... more In this paper, we introduce a lexicon-based method for identifying political topics in social media messages. After discussing several critical shortcomings of unsupervised topic identification for this task, we describe the lexicon-based approach. We test our lexicon on candidate-generated campaign messages on Facebook and Twitter in the 2016 U.S. presidential election. The results show that this approach provides reliable results for eight of nine political topic categories. In closing, we describe steps to improve our approach and how it can be used for future research on political topics in social media messages.
International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation, 2017
To understand political campaign messages in depth, we developed automated classification models ... more To understand political campaign messages in depth, we developed automated classification models for classifying categories of political campaign Twitter and Facebook messages, such as calls-to-action and persuasive messages. We used 2014 U.S. governor’s campaign social media messages to develop models, then tested these models on a randomly selected 2016 U.S. presidential campaign social media dataset. Our classifiers reach .75 micro-averaged F value on training sets and .76 micro-averaged F value on test sets, suggesting that the models can be applied to classify English-language political campaign social media messages. Our study also suggests that features afforded by social media help improve classification performance in social media documents.
#SMSociety17 Proceedings of the 8th International Conference on Social Media & Society, 2017
To date, little attention has been paid to the temporal nature of campaigns as they respond to ev... more To date, little attention has been paid to the temporal nature of campaigns as they respond to events or react to the different stages of a political election – what we define as strategic temporality. This article seeks to remedy this lack of research by examining campaign Facebook and Twitter messaging shifts during the 2016
U.S. Presidential general election. We used supervised machine learning techniques to predict the types of messages that campaigns employed via social media and analyzed time-series data to identify messaging shifts over the course of the general election. We also examined how social media platforms and candidates’ party
affiliation shape campaign messaging. Results suggest differences exist in the types of campaign messages produced on different platforms during the general election. As election day drew closer, campaigns generated more calls-to-action and informative
messages on both Facebook and Twitter. This trend existed in advocacy campaign messages as well, but only on Twitter. Both advocacy and attack tweets were posted more frequently around Presidential and Vice-Presidential debate dates.
In Proceedings of the 2015 Association for Computing Machinery iConference. , 2015
This study examines academic opinion expressions in citation context. We first developed an annot... more This study examines academic opinion expressions in citation context. We first developed an annotation schema to annotate three aspects of each academic opinion expressed in a citation statement: rhetorical purpose, content aspect, and opinion polarity. We then annotated two samples: a natural science sample consisting of biomedical journal articles, and an engineering sample consisting of conference papers in the natural language processing field. A comparison of the annotations on the two samples showed disciplinary differences in citation opinion expressions. The result contributes to the understanding of academic opinion expressions in citation context and the development of automated citation opinion analysis tools to assist researchers' literature search and navigation.
Keywords: Citation Analysis; Opinion Mining; Natural Language Processing; Biomedicine
Paper presented at the National Communication Association annual conference, Chicago, Illinois., 2014
This study examines the impact of the intensity use of Sina Weibo, the most popular social networ... more This study examines the impact of the intensity use of Sina Weibo, the most popular social network site in China, on users’ online political expression and social capital. The study also explores whether online bridging and bonding social capital in Weibo can predict political expression. In addition, the mediating impact of online bridging social capital on the relationship between the intensity of Sina Weibo use and political expression is investigated. Structural equation modeling analysis is conducted to analyze a web-based survey data collected from 306 Sina Weibo users. The results indicate that a positive association exists between the intensity of Sina Weibo use and social capital. Online bridging social capital predicts, rather than bonding social capital, political expression but with marginal effect. The intensity of Sina Weibo use has both direct and indirect relationship with political expression, and bridging social capital serves as a mediator in the indirect relationship.
Keywords: Sina Weibo, social network sites, social capital, political expression
Proceedings of the 8th International Conference on Social Media & Society, 2017
In this paper, we introduce a lexicon-based method for identifying political topics in social med... more In this paper, we introduce a lexicon-based method for identifying political topics in social media messages. After discussing several critical shortcomings of unsupervised topic identification for this task, we describe the lexicon-based approach. We test our lexicon on candidate-generated campaign messages on Facebook and Twitter in the 2016 U.S. presidential election. The results show that this approach provides reliable results for eight of nine political topic categories. In closing, we describe steps to improve our approach and how it can be used for future research on political topics in social media messages.
International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation, 2017
To understand political campaign messages in depth, we developed automated classification models ... more To understand political campaign messages in depth, we developed automated classification models for classifying categories of political campaign Twitter and Facebook messages, such as calls-to-action and persuasive messages. We used 2014 U.S. governor’s campaign social media messages to develop models, then tested these models on a randomly selected 2016 U.S. presidential campaign social media dataset. Our classifiers reach .75 micro-averaged F value on training sets and .76 micro-averaged F value on test sets, suggesting that the models can be applied to classify English-language political campaign social media messages. Our study also suggests that features afforded by social media help improve classification performance in social media documents.
#SMSociety17 Proceedings of the 8th International Conference on Social Media & Society, 2017
To date, little attention has been paid to the temporal nature of campaigns as they respond to ev... more To date, little attention has been paid to the temporal nature of campaigns as they respond to events or react to the different stages of a political election – what we define as strategic temporality. This article seeks to remedy this lack of research by examining campaign Facebook and Twitter messaging shifts during the 2016
U.S. Presidential general election. We used supervised machine learning techniques to predict the types of messages that campaigns employed via social media and analyzed time-series data to identify messaging shifts over the course of the general election. We also examined how social media platforms and candidates’ party
affiliation shape campaign messaging. Results suggest differences exist in the types of campaign messages produced on different platforms during the general election. As election day drew closer, campaigns generated more calls-to-action and informative
messages on both Facebook and Twitter. This trend existed in advocacy campaign messages as well, but only on Twitter. Both advocacy and attack tweets were posted more frequently around Presidential and Vice-Presidential debate dates.
In Proceedings of the 2015 Association for Computing Machinery iConference. , 2015
This study examines academic opinion expressions in citation context. We first developed an annot... more This study examines academic opinion expressions in citation context. We first developed an annotation schema to annotate three aspects of each academic opinion expressed in a citation statement: rhetorical purpose, content aspect, and opinion polarity. We then annotated two samples: a natural science sample consisting of biomedical journal articles, and an engineering sample consisting of conference papers in the natural language processing field. A comparison of the annotations on the two samples showed disciplinary differences in citation opinion expressions. The result contributes to the understanding of academic opinion expressions in citation context and the development of automated citation opinion analysis tools to assist researchers' literature search and navigation.
Keywords: Citation Analysis; Opinion Mining; Natural Language Processing; Biomedicine
Paper presented at the National Communication Association annual conference, Chicago, Illinois., 2014
This study examines the impact of the intensity use of Sina Weibo, the most popular social networ... more This study examines the impact of the intensity use of Sina Weibo, the most popular social network site in China, on users’ online political expression and social capital. The study also explores whether online bridging and bonding social capital in Weibo can predict political expression. In addition, the mediating impact of online bridging social capital on the relationship between the intensity of Sina Weibo use and political expression is investigated. Structural equation modeling analysis is conducted to analyze a web-based survey data collected from 306 Sina Weibo users. The results indicate that a positive association exists between the intensity of Sina Weibo use and social capital. Online bridging social capital predicts, rather than bonding social capital, political expression but with marginal effect. The intensity of Sina Weibo use has both direct and indirect relationship with political expression, and bridging social capital serves as a mediator in the indirect relationship.
Keywords: Sina Weibo, social network sites, social capital, political expression