Knowledge Graphs (KGs) are now essential tools for organizing, linking, and exploiting information. However, their automatic construction from unstructured texts remains a major scientific challenge, particularly when dealing with heritage corpora such as historical dictionaries and encyclopedias (Hofer et al. 2024). These documents represent rich and invaluable sources, yet their exploitation is hampered by linguistic variability, the lack of explicit structure, and the complexity of historical references. In this context, geographical data —place names, spatial and hierarchical relations— are of particular importance, both for the humanities and for artificial intelligence (Bhandari et al. 2023).