L’applicazione di reti neurali convoluzionali per l’estrazione automatica della conoscenza dei beni culturali da archivi multimediali

Abstract

The increasing application of Artificial Intelligence (AI) technologies is radically transforming the ways of analyzing multimedia files, affecting the possibilities of retrieving hidden information. These opportunities are much more evident in the field
of cultural heritage, wherein the assets are characterized by a heterogeneity of elements that can be well-documented and described in archival documents. The present article aims to highlight the opportunities for applying convolutional neural networks (CNNs) on archival documents in the form of images or audiovisuals related to cultural heritage. The work aims to theoretically identify the main techniques based on CNNs to explore the content of these types of documents to automatically extract key information related to cultural heritage useful to promote knowledge discovery and dissemination services.