Digitization is making ever larger data sets available to researchers. Qualitative research is also benefiting from AI, which is intended to support researchers and archivists in their work and at the same time be tailored to their needs. A quantitative quantum leap in new analysis techniques not only accelerates research processes, but also opens up entirely new questions of a qualitative nature. In the end, we want to show that it is possible to "interrogate" a cabinet (i.e. digitized archival data) and that the cabinet (i.e. a AI large language model (LLM) trained on archival data) can "interrogate" itself in the sense of delivering meaningful answers cross-referenced to the original files.  The cabinet’s (i.e. the AI LLM’s) general capability to generate linguistically meaningful and reasonable “answers to the questions" relies on pre-trained models, whereas the exact content of entire archives made available to it in the final training or fine-tuning steps makes it possible to ask extremely archive-specific questions. To improve this technology, we will develop three very heterogeneous "digital filing cabinets", i.e. based on the same algorithms and software, in parallel, in close collaboration with three established and prestigious archival institutions.

Alexander Ermakov and Esther Gardei won the #Hackathon2024 against Antisemitism! This prize was the first step in our joint interrest in developing an AI application with a question function for anti-Semitism research, the ‘Speaking Filing Cabinet.’ This application accesses thousands of data on the topics of anti-Semitism and ‘Aryanisation’, which will be made accessible to researchers by archives in 2024 with the aim of managing the unmanageable amount of data.

The project is a first step in connecting the research area of the BCDH and the BCRS in the field of AI.