Use of Large Language Models for location detection on the example of the terrorism and extremism event database

This technical report discusses the potential use of Large Language Models for location detection with a focus on the JRC Terrorism and Extremism Database. The report highlights the current inaccuracies in the database's location detection algorithm, which uses adhoc created embeddings, struggles with contextual issues, and has difculty with translated location names. These issues can lead to misclassifed events, negatively impacting the quality of the tool. The report suggests exploring more accurate and sophisticated approaches, taking advantage of recent advancements in Artifcial Intelligence, to increase the accuracy of the automated classifcation thus helping reducing human intervention.