The TALOS Lab is proud to share that Fotini Koidaki, PhD Candidate in RT3: Corpus Analysis, successfully presented her work at the 17th International Conference on Greek Linguistics (ICGL17), held at the University of Cambridge (UK) from 23–26 September 2025.

Her talk, titled “From Page to Pixel: The 19th Century Greek 8.0 OCR Model”, introduced the development and applications of an Optical Character Recognition (OCR) model tailored for 19th- and early 20th-century Greek prints.

Developed as part of her PhD research within TALOS-AI4SSH, the “19th Century Greek 8.0” model was trained on 894 authentic pages using the Transkribus platform and achieved an outstanding Character Error Rate (CER) of 0.9. It stands as the first publicly accessible model in a series designed for digitizing Greek printed heritage, offering scholars a reliable and open resource for Digital Humanities research.

During her presentation, Fotini discussed:

  • The process of creating and training the model and the key factors influencing its performance,
  • Practical steps for users to access and apply it through Transkribus,
  • Future improvements and personalization options for broader digitization initiatives.

Her contribution highlighted how open AI tools can significantly advance the digitization, accessibility, and study of historical Greek texts, bridging technology and philology in innovative ways.

🔗 Conference website: ICGL17 – University of Cambridge
📄 Read the full abstract: ICGL17 Abstracts (PDF)
🧠 Explore the OCR Model: 19th Century Greek 8.0 on Transkribus