Topic modeling method for analyzing social actor discourses on climate change, energy and food security

Abstract

Debates on climate change, energy and food security concerns have underscored the need for new methods and tools to explore and understand the complexity of these relevant issues. In this study, we used unsupervised probabilistic modeling—latent Dirichlet allocation (LDA)—to examine the changes in social policy debates related to ethanol production in Brazil and its relationship with climate change and food security. We analyze a large amount of data obtained from Brazilian newspapers, government and business documents, and the bulletins of nongovernmental organizations and social movements from 2007 through 2017. The results from the LDA application allowed us to identify key topics, detect novel trends, and follow them through time, in addition to exhibiting the limitations encountered in identifying social actor discourses. To overcome these limitations, we combine LDA and discourse analysis to enable the construction of topics, concepts, and discourses of the actors surrounding the issue of this study. The findings verify the utility of these techniques by emphasizing the themes and discourses of the actors involved and by identifying the determinant positions of the Brazilian actors in the discussions on ethanol production and its competition with food security and the contributions of ethanol as a renewable energy source to mitigate climate change. This finding provides insights for water–energy–food nexus research.