The Foggia Occupator Corpus: Digitisation, Annotation, and Computational Analysis of an Occupation‑Era Newspaper (1945-1946)

Abstract

Historical newspapers are crucial sources yet often remain undigitised or lack machine-readable text. We present the Foggia Occupator corpus, a linguistically enriched, openly licensed resource built from twenty-two issues (Dec 1945–Aug 1946) of a weekly newspaper produced by U.S. personnel in occupied Foggia, Italy. High-resolution scans were processed via OCR with LLM‑assisted correction (GPT‑4o) and full human verification, then segmented into 874 articles ( 216k tokens). We annotate topics, named entities and typed relations via a semi‑automatic pipeline with manual reconciliation, and perform argument mining on civics‑ and conflict‑related content, yielding 1,735 arguments. The entity–relation layer supports network analyses that reveal sparse, modular structures linking military units, civic bodies, and social life. We release TEI‑XML with entity spans, JSON article files with metadata, CSVs of entities/relations with temporal counts, and an arguments JSON, all under a Creative Commons 4.0 licence. Beyond documenting an in‑between moment of reconstruction, the resource enables benchmarking for OCR‑robust NER/RE and studies of framing, stance, and community structure in post‑war local media.

Resources

Details