Applying (semi-)automatic metadata to early modern normative texts. Annif and <i>Policeygesetzgebung</i> from the City-State of Bern (1528–1798)

Abstract This study investigates the application of modern digital tools to analyze handwritten normative texts from the City-State of Bern (1528-1798). By leveraging the Simple Knowledge Organization System (SKOS) and automatic metadata generation tools such as Transkribus for Handwritten Text Recognition (HTR) and Annif for automated subject indexing, we aim to enhance the efficiency and accuracy of historical text analysis. Our methodology involves digitizing 4,550 police ordinances, manually segmenting them, and employing a hierarchical SKOS framework with 1,830 distinct labels (Rhonda Organisation 2024). Annif’s backends, including TF-IDF, Umikuji Parabel, Umikuji Bonsai, and Machine Learning Language Model, were evaluated using Normalized Discounted Cumulative Gain (NDCG) metrics to generate relevant metadata. The results demonstrate that the Umikuji backends achieved high NDCG scores, highlighting their effectiveness. This research highlights the transformative potential of integrating digital tools in the humanities, providing a scalable solution for managing large historical datasets and enhancing access to historical texts. The findings advocate for broader adoption of these technologies in historical research, promoting greater collaboration and innovation. Future research will explore the application of these methods to other historical corpora and additional backend systems.