Ancient dictionaries and encyclopedias, such as Diderot and d'Alembert's Encyclopédie (1751–1772), contain knowledge—particularly geographical information—that is invaluable for studying how such knowledge has evolved over the past centuries. However, the thousands of articles in these works require automated tools to extract reliable and structured information. Our tool is based on models specifically trained on historical texts and it enables the construction of an RDF graph representing geographical entities and their spatial relationships. An interactive prototype1, as well as the trained models2, are available on Hugging Face, and the source code is available on GitLab3.