Discourse network analysis is a toolbox of research methods for the analysis of actor-based debates, such as policy debates or political discussions. Examples include the policy debates on climate change, pension politics, or around the introduction of large infrastructure projects. Political actors typically include organisations (interest groups, political parties, government agencies etc.) or individual persons (legislators, celebrities etc.). These actors make statements about policy instruments, solution concepts, narratives, frames, issues, arguments etc. in the media or other arenas, and these statements are temporally and cross-sectionally interdependent. Actors build coalitions in a debate by reinforcing each other’s statements or making similar statements, and they frequently contradict each other over time among these coalitions. The goal of discourse network analysis is to explore, describe, and draw inferences about these processes based on text data and based on a manual qualitative coding in combination with quantitative social network analysis.
The short course will first introduce a few examples from the literature on discourse networks, define the key concepts, and discuss theoretical frameworks that are compatible with the methodological approach of discourse network analysis. We will then consider different text sources, different types of debates, and obstacles in the coding process. The software Discourse Network Analyzer (DNA) 2.0 will be introduced in a hands-on computer lab session, and after demonstrating the coding process in this software, we will proceed with the analysis of the resulting network data in external network analysis software packages, such as visone. The next theoretical block will cover the different export options and algorithms available to the user when exporting network data from DNA. Different data transformations are applicable depending on the type of debate, the nature of the data-generating process, and the goals of the analysis. We will consider, using DNA and network visualisation software like visone, how the choice of the algorithm or method leads to different results, why that is the case, and what method to choose in a given situation. We will also cover basic network analysis techniques that are useful for the analysis of discourse networks (such as community detection algorithms and centrality), and we will briefly consider their implementation in software packages. The course will then proceed by introducing rDNA, which is a package for the statistical computing environment R that permits the user to import network data resulting from DNA directly into R. We will discuss several best practices for doing cluster analysis and other procedures with the data in R, and we will discuss options for the temporal analysis of coalitions and other key features of a debate over time. Finally, we will briefly introduce the statistical or inferential analysis of temporal discourse network data using relational event models for bipartite signed graphs as implemented in the R package rem, and we will discuss the data requirements and theoretical insights to be gained from such an inferential analysis.
The course will be primarily based on lectures and lab tutorials, but participants will be given the opportunity to discuss their own projects and work with their own data in the tutorial lab sessions. To do so, participants should have their text data ready in machine-readable form and bring their own laptops. The course is designed as an introductory short course, but will cover some advanced topics like statistical analysis in R and network analysis in visone. While existing skills in these domains would be an advantage, the course will introduce these skills as far as possible also to a lay audience. To follow the parts that focus on R, however, basic familiarity with R will be required.
Note that this course is neither an introduction to the general principles of qualitative inquiry or content analysis, nor an introduction to quantitative text analysis or machine learning. The focus of the course is specifically on the methodological toolbox of discourse network analysis, from the conceptual stage through manual coding up to inferential network analysis of discourse network data.