Use of Large Language Models to Automate the Extraction of Interrogation Data within Joint Investigation Teams

The interdisciplinary research combines insights from criminal justice and information technology, utilising Graph Neural Networks and Retrieval-Augmented Generation and advanced prompting techniques to enable complex analysis of relationships between entities, such as phone numbers, emails, accounts, people or vehicles, which would improve investigators’ ability to process and interpret large volumes of heterogeneous data. Joint investigation teams face significant challenges in processing and analysing large multilingual files. The basic forensic method is interrogation, along with preparation and execution. This problem is addressed in the context of a research task that aims to create a system using artificial intelligence with large-scale language models. Thus, the article presents the readers with the possibilities of solving the research task in the implemented project, which is aimed at creating an information system to facilitate and streamline the work of police forces within joint investigation teams.