Study reveals major blind spot in OCR systems: Black historical newspapers and marginalized archives systematically excluded from training and evaluation

OCR evaluation methods are heavily biased toward modern, Western, and institutional documents, masking how systems perform on historical materials

Comprehensive PRISMA framework review of OCR papers from 2006-2025 found Black newspapers and community-produced historical documents rarely appear in training data or benchmarks

Current vision transformer and multimodal OCR systems fail to account for challenges in historical documents, including degraded materials, non-standard typography, and complex layouts

The invisibility of marginalized archives in OCR evaluation perpetuates inequities in digital preservation and accessibility of historically important community records