Tools3h ago
FineBooks Benchmarks OCR Models For Training Data
The dots.mocr model achieved 97.6 percent character accuracy across 2,000 historical book pages. Developed by Hugging Face and EleutherAI, the project identifies a cost-effective path to cleaning legacy text at under two dollars per thousand pages. While insufficient for scholarly work, the accuracy suffices for LLM training. This solves a persistent data quality bottleneck.