The historical geography of crime: analyzing Istanbul sharia court records with artificial intelligence

The thesis reconstructs the historical crime geography of Ottoman Istanbul by combining AI-assisted analysis of Sharia Court Registers with Historical Geographic Information Systems (HGIS). The dataset comprises 2,653 crime records extracted—via AI-based preprocessing—from the entire corpus of 54,593 rulings contained in 100 Istanbul court volumes available on kadisicilleri.org. Using Python pipelines and OpenAI-powered NLP (NER and thematic coding), unstructured texts were transformed into structured data; outputs were then manually verified for accuracy. The records were mapped as GIS layers in ArcGIS Pro to examine spatial clustering and density patterns across offense types. Findings indicate elevated risk per area in the administrative-commercial core (within the walled city) and along the Tophane–Golden Horn shipyard–market corridor. Crimes against religion/public order—especially alcohol-related cases—concentrate in the port hinterland, while crimes against bodily integrity show high densities in the Istanbul Court and Tophane; due to its large jurisdiction, Üsküdar exhibits high absolute counts but low density. Corruption-related crimes (malfeasance/bribery) cluster around the Bâb-ı Âli administrative core, underscoring bureaucratic centrality. Overall, a "core–coastal corridor–periphery" hierarchy is reproduced across crime categories, aligning with Routine Activity and crime-of-place/hot-spots perspectives. Methodologically, the study demonstrates that classical historical sources can be analyzed systematically, rapidly, and reproducibly through the integration of AI/NLP and HGIS.