This paper presents BELHISFIRM, a digital humanities project that transforms over two centuries of Belgian corporate records from the Belgian Official Gazette into structured, machine-readable research data. The project builds a six-stage pipeline covering digitization, automatic text recognition, layout classification, named entity recognition, record linkage, and linked data publication to convert millions of archival fragments into coherent firm histories and networks. A key methodological focus is the evaluation of vision-language models versus traditional OCR for handling the complex, irregular layouts of historical company records. The resulting data is published as linked data on an interactive SAMPO-UI platform, enabling researchers to query, browse, and visualize long-term corporate trajectories in Belgian business history.
Vercruysse et al. (Wed,) studied this question.