Look who’s talking: Two-mode networks as representations of a topic model of New Zealand parliamentary speeches

Abstract

Quantitative methods to describe the participation to debate of Members of Parliament and the parties they belong to are lacking. Here we propose a new approach that combines topic modeling with complex networks techniques, and use it to characterize the political discourse at the New Zealand Parliament. We implement a Latent Dirichlet Allocation model to discover the thematic structure of the government’s digital database of parliamentary speeches, and construct from it two-mode networks linking Members of the Parliament to the topics they discuss. Our results show how topic popularity changes over time and allow us to relate the trends followed by political parties in their discourses with specific social, economic and legislative events. Moreover, the community analysis of the two-mode network projections reveals which parties dominate the political debate as well as how much they tend to specialize in a small or large number of topics. Our work demonstrates the benefits of performing quantitative analysis in a domain normally reserved for qualitative approaches, providing an efficient way to measure political activity.

How has Science Education changed over the last 100 years? An analysis using natural language processing

“…As of writing, LDA has been developed and used for close to two decades, and has been applied to a variety of fields and applications. For example, researchers in political science have used it to extract the key themes and political priorities from large numbers of transcribed political speeches (Curran et al, 2018; Grimmer & Stewart, 2013) and to quantify the degree of polarization in political commentary (Balasubramanyan et al, 2012; Roberts et al, 2016). In the software industry, LDA has been used to analyze large software databases to try comprehend the functions of different parts of the code (Maskeri et al, 2008).…”

Combining topic models with bipartite blockmodelling to uncover the multifaceted nature of social capital

“…Consequently, the probabilities in the matrix are not randomly distributed and cannot be used for a Gaussian bi-clustering procedure. According to the criterion adopted by Curran et al [37], we dichotomised the probabilities in each column by assigning a value of one if the probability was twice as high as the expected loading when the topics were equally important for an article (i.e., a loading of (1/13) � 2). The resulting matrix, which served as the basis for the bi-clustering procedure, had 11,975 rows (articles) and 13 columns (topics).…”

“…This helps with the identification of gaps in the literature and, therefore, opportunities for future research; and (b) we demonstrate the potential of combining topic models with bipartite blockmodelling, a technique stemming from network analysis. Topic models have been employed on scientific publications before [33][34][35][36]; combining network analysis with topic models is also fairly common [37][38][39]. However, the network properties analysed in these studies are usually metadata related to the documents, e.g., whether publications cite one another [38,39].…”

Using parsed and annotated corpora to analyze parliamentarians' talk in Finland

“…Parliament is a crucial research interest for many disciplines ranging from politics, history, and sociology to the study of rhetoric. Parliaments have for a long time been the main object of struggles for democratization (Bonin, 2020;Ihalainen, Ilie, & Palonen, 2016;Kurunmäki, Nevers, & te Velde, 2018) and have also received attention from a digital humanities perspective (see, e.g., Curran et al, 2018;Grimmer, 2013). Today, parliament is the main forum for the nationwide political debate and has a formalized link to political decision making in democratically governed countries (e.g., Ilie, 2010).…”