Social Network Analysis Tools for Understanding the Evolution of Twitter Conversations

• In Event: Social Network Tools for Twitter Analysis

Abstract

Twitter is a popular source of data for researchers who are interested in studying public conversations. Typically, Twitter data are used to capture the prevalence of keywords indicating the importance of a topic. But of course, Twitter is fundamentally a social medium. More than just indicating a societal conversation, Twitter data can reveal the pattern of ties among users that emerges around one or a set of topics.
The objective of this paper is to demonstrate how social network analysis tools can be used with data scraped from Twitter to gain a deeper understanding of these conversations. A discussion on Twitter can be viewed as a social network in which links are formed when users mention, reply to, retweet, or quote each other. These interactions may lead to changes in the conversation over time. Depending on the context, researchers could be interested in changes in sentiment, language used, or some other aspect of the conversation.
Twitter data is typically scraped using a list of keywords related to the conversation of interest. We were interested in identifying some of the key issues motivating the conversation around a school board recall in Loudoun County, Virginia in June 2021. First, we scraped tweets that included any of the following combinations of keywords: “Loudoun school board”, “Loudoun recall”, “LCPS school board”, or “LCPS recall”. Then, tweets were coded for key issues: critical race theory (CRT), COVID, LGBTQ, school shootings, or none of these. As shown in Figure 1, there was an increase in tweets about the school board recall around June 11 mostly related to CRT and a second, larger increase around June 25 driven by concerns about CRT and LGBTQ issues.
Dynamic network visualizations are especially useful when trying to understand how Twitter conversations change over time. We used a dynamic animation to visualize daily changes in the school board recall network over the month of June. In Figures 2-4, snapshots from the animation show the state of the network on three days. Information from user profiles and tweets can be viewed by clicking on nodes and links, respectively. Twitter details can also be used to inform the color, shape, and size of nodes and links to highlight attributes that are of particular interest. This type of animation can be useful for tracking when new language or ideas are introduced into the conversation and how they spread through the network. Dynamic animations and other methods used in social network analysis provide powerful tools for understanding the evolution of Twitter conversations.

To view the full dynamic animation for the Loudoun County school board recall network:
https://www.dropbox.com/s/i3xr5mqwmqi5tkk/loudoun%20co%20animation.html?dl=0

Note. The controls in the upper right corner can be used to adjust the speed of the animation and the mouse scroll wheel can be used to zoom in and out. Clicking on nodes and links to view additional attributes works best when the animation is paused and not transitioning. Pausing between transitions is much easier if the speed of the animation is slow.

Authors

• Nicole Jess, Michigan State University
  Presenting Author

• Hasan Gokberk Bayhan, Michigan State University
  Non-Presenting Author

• Yi-Chih Chiang, Michigan State University
  Non-Presenting Author

• Shimeng Dai, Michigan State University
  Non-Presenting Author