The project aims to collate and analyse scientific publications on the risks and impacts of local, regional and large-scale regime shifts for Earth resilience, ecosystems and social systems. The objective is to utilise machine learning to explore the potential of topic modelling methods for classifying text chunks from relevant scientific publications. The outcomes will inform improvements to the regime shifts database and other scientific literature analysis projects hosted at the Stockholm Resilience Centre/Stockholm University, which will be accessible to all staff.