The preservation of architectural heritage and the landscape is an essential part of the conservation and enhancement of our cultural heritage. Achieving this goal requires in-depth knowledge of all aspects of the structure to be preserved: from its shape to the materials used; from the construction techniques employed to its relationship with its surroundings and the historical and cultural background in which it is located. In this context, archives are a primary source of knowledge and yet are often difficult to access and inadequately indexed, placing significant limitations on consultation and dissemination of information among researchers and interdisciplinary working groups. Modern OCR (Optical Character Recognition) techniques can facilitate this process. However, their application to historical documents, especially manuscripts, is still challenging and costly.
This paper presents the preliminary results of the application of CHURRO, an open-weight artificial intelligence model for the transcription of historical texts, to a corpus of manuscript documents relating to the Geirato siphon bridge, part of Genoa’s historic aqueduct, demonstrating the feasibility of a cost-effective, replicable and locally executed digitisation pipeline.
