Skip to main content

Exploring new data sources with ChatGPT and Co

Participants of the first Text-as-Data workshop hosted by the University of Potsdam
Photo: Deva Charan Jarajapu

Imagine you're trying to understand public opinion on climate change, how people deal with a heatwave, or what impact droughts have on agriculture, forestry, tourism, or energy systems. You could conduct surveys to find out, or you could make use of information that is already there. Newspapers, social media, government reports, and scientific articles all write a lot about these issues and offer a wealth of information at our fingertips. With the increasing risk of natural hazards due to climate change, we need to use all the information that we have to find solutions. Text-as-data is the name for a host of methods (for example, language models, like ChatGPT) that analyse unstructured text data to extract meaningful information and patterns.

The community of experts applying text-as-data methods in hydrology, climate change and natural hazards research is small, but growing. In March, the first workshop on “Text-as-data in Hydrology, Natural Hazards, and Climate Change Research: exploring challenges and opportunities” hosted by the University of Potsdam’s Analysis of Hydrologic Systems Group (Dr. Lina Stein) and co-organised by the Helmholtz Centre for Environmental Research (Dr. Mariana De Brito) and the Potsdam Institute for Climate Change (Jakob Lochner), brought together a wide variety of researchers for a day of networking and interdisciplinary discussion.  

After two keynote talks focusing on recent technical developments, it was time for the attendees to discuss. What opportunities are there to utilise text-as-data in the environmental sciences? What barriers or bias do we have to address to make these opportunities come true? And what are the real world steps, that the text-as-data community might take to further develop their field?

Overall, the workshop served as a valuable opportunity for scientists to collaborate and explore the vast potential of text data analysis in addressing pressing societal issues related to water resources, climate variability, and disaster risk management. Future meetups are already in planning including the next workshop for spring 2025 which will take place at the Helmholtz Centre for Environmental Research, Leipzig.

Participants of the first Text-as-Data workshop hosted by the University of Potsdam
Photo: Deva Charan Jarajapu