Publication of a paper

The paper “Automated information extraction from text variables in event datasets with large language models,” co-authored by Laura Braun and Christian Oswald, demonstrates how large language models (LLMs) can enrich existing conflict event data. Focusing on abduction and forced disappearance events in the Armed Conflict Location and Event Data (ACLED), they show how variables such as the exact number of abductees, their gender, if they were underaged, and whether a ransom was demanded can be extracted
accurately and at scale. This approach enables near-real-time monitoring of incidents such as government kidnappings and large-scale abductions of minors. The authors also illustrate that open-weight models perform just as well as closed-weight ones, making these advances broadly accessible for both researchers and practitioners.

The paper can be found here and will be presented at the MZESDVPW "Methods of Political Science" Conference in Mannheim, Germany, in late March and the COMPTEXT conference in Vienna, Austria, in late April 2025.

Picture: © CCEW