AI Assistant for Heritage Knowledge Management: Retrieving Integrated Data from Ontology-Driven HBIM and Archival Documents

Managing heritage assets nowadays requires systems capable of han dling the complexity of diverse data types, from structured Heritage Building Information Models (HBIM) and information ontologies to historical and archival documents and investigation reports. Current practices struggle to combine these diverse types of information and formats, limiting the possibility of comprehensive heritage interpretation and reasoning processes. This paper introduces an AI Assistant to integrate an ontology-driven HBIM with textual documentation for natural language querying, inference, and knowledge management. The system consists of (1) IndArch ontology used to structure representation templates in the (2) HBIM model, where the parameters, such as spatial, structural, and historical event-based features, are extracted and converted into JSON format for compatibility with the AI model; (3) unstructured data, including a variety of documentation (pdf files) are directly processed through (4) the AI Assistant’s backend, which leverages “file search” and “function calling” capabilities to query and cross-reference information. The proposed framework, applied to the Sanctuary of Hercules and the former Segrè Papermill in Tivoli, Italy, reconstructs the room functions and the latest papermill production process; the AI Assistant demonstrates its ability to infer knowledge from structured and unstructured sources, enhance accessibility, and reduce technical barriers for heritage professionals. Finally, the system is qualitatively evaluated to assess model performance and identify areas for improvement in the AI training phase