Digitalisation Workflows Using Transformer Models

This research article presents a case study focusing on the application of transformer-based technologies in the digitalisation workflows of cultural heritage data. The integration of transformer models, such as BERT and GPT-4, has significantly reshaped the processes involved in the digitisation of cultural heritage (CH) assets. The paper outlines two specific projects, namely 'Themenportal Wiedergutmachung' and 'Deutsche Digitale Bibliothek', demonstrating how these models enhance efficiency and improve the quality of data processing within CH initiatives. The transformative impact of these technologies is discussed, detailing both the advantages of increased accuracy in data interpretation and potential challenges such as data privacy concerns. The research also examines the implications of employing these advanced AI solutions across various metadata and unstructured data sources, highlighting the necessity for careful consideration of the balance between technological benefits and ethical responsibilities. Overall, the document provides insight into the evolving landscape of CH digitalisation driven by innovative AI methodologies.