Mapping actor and issue networks in urban sustainability over time using automated content analysis of media data

To be submitted for consideration to the panel ”From Texts to Networks: Semantic, Socio-Semantic, and Discourse Networks” --------------------- The UN Agenda 2030 and its associated sustainable development goals (SDGs) put an emphasis on the role of cities for achieving a sustainable future for all. Implementing the SDGs in cities is a complex governance challenge, touching on issues ranging from gender equality over biodiversity to climate protection. This means that implementing the SDGs requires cross-scale governance involving solutions in many policy areas and all societal sectors. Actor networks play a key role for successful cross-scale governance in complex sustainability governance settings. These networks, also called governance networks, consist of organizations from all societal sectors, including civil society, the private sector, government and scientific institutions. In urban contexts, the need to consider actor networks is especially pronounced given the extraordinary diversity and density of actors encountered in urban governance arenas. For research on urban sustainability governance networks to be useful in supporting actor networks for effective cross-scale governance, it needs to find ways to reliably map and understand these networks. Empirically mapping rapidly evolving governance networks in urban sustainability governance is a complex task. Still, a new research frontier, using the possibilities of data sources and tools arising from digitalization, is increasingly offering ways to handle this complexity and to map governance networks through the use of natural language processing, enabling automated text analysis. In this paper, I introduce the foundations for a data processing pipeline mapping actor networks involved in urban sustainability governance using media data. The processing pipeline uses supervised machine learning to first categorize text at the paragraph level, assigning paragraphs to SDG implementation areas (such as mobility). In a second step, we train a named entity recognition classifier to identify organizational actors occurring in these paragraphs and then build on a recently proposed model to extract narratives from large text corpora to categorize actor activity at the sentence level. In a third step, we then use unsupervised text classification to identify governance issues in the implementation area (such as district level implementation of bike infrastructure) in a bottom-up approach by grouping bundles of related actor activity. The final output is a time-stamped bipartite actor-issue graph relating actors to governance issues and through its projection also relating actors and issues to each other. I demonstrate the processing pipeline and its use in mapping urban sustainability governance networks by applying it to a dataset of news articles from the city Zurich containing 104’000 paragraphs from media articles between 2010 and 2021 occurring in the local (Zurich) section of two national-level, Zurich based newspapers.