This page needs JavaScript! Please enable it to continue.

This website uses JavaScripts. If you use an adblocker, content may not be displayed or may not be displayed correctly.

Achim Rabus: Beyond Transkribus – exploring next‑generation Handwritten Text Recognition technologies

Lecturer(s)Achim Rabus
Contact personAchim Rabus
Emailachim.rabus@slavistik.uni-freiburg.de
DateWednesday, 5th November 2025, 10:00 - 11:30
LocationDigital Humanities Lab, KG I, Raum 1026 Freiburg Germany
Save this event iCal

Beyond Transkribus – exploring next‑generation Handwritten Text Recognition technologies

Achim Rabus

 

Historical corpus linguistics and scholarly editing need large amounts of digitized and transcribed sources in many languages. In recent years, Handwritten Text Recognition (HTR) has become central, with Transkribus as a powerful and user‑friendly – but proprietary and paid – market leader. This work‑in‑progress talk gives an overview of current free and commercial alternatives, from compact and efficient models such as kraken and PyLaia to transformer‑based approaches like TrOCR, and to Large Language Models and Visual Language Models like (open-weights) Qwen3‑VL or (commercial) Gemini. I discuss strategies and implications for training and fine‑tuning these models for different languages and scripts, as well as inference, and I show first application results. I also present an early prototype, built with the promptotyping method, of a transcription tool that lets users choose among different HTR engines.

 

And via Zoom.