Automated Text Recognition
Automated Text Recognition
A teaching resource on OCR and HTR for historical documents — from classical pipelines and open-source tools to modern transformer and vision-language models.
What you will find here
Introduction
What is OCR and HTR? The recognition pipeline, key metrics (CER/WER), and the challenges of historical documents.
eScriptorium
The open-source annotation and transcription platform built on Kraken. Workflow, training, export, and further resources.
Open-Source Models
A curated overview of publicly available Kraken/eScriptorium models for medieval and early modern manuscripts.
HTR-United
The community initiative for sharing HTR training data and models under standardized metadata.
TrOCR
Microsoft’s transformer-based OCR model — architecture, pre-training, fine-tuning, and when to use it.
Vision Language Models
GPT-4o, Gemini, and open VLMs applied to historical text recognition — possibilities and limitations.
Quiz
Test your knowledge with a Kahoot-style multiple-choice quiz covering all sections of this resource.
Literature
Key readings from the Zotero group Automated Text Recognition, searchable and organized by year.
Workshops
Schedules and materials for hands-on ATR workshops. Next: DMSI 2026 Kalamazoo, 13 May 2026.