How can historical newspapers and radio broadcasts 📰📻 from 500+ sources, 20+ institutions, multiple languages, and decades of digitization be represented for ML, information retrieval, and archival fidelity? At PressMint / LREC2026, Pauline Conti, Simon Clematide, and Maud Ehrmann presented a framework now covering 567 titles, 73M+ content items, and 24B+ tokens 📊 🔗 Read the full paper: https://infoscience.epfl.ch/handle/20.500.14299/263437