HTR/ATR Infrastructure and GLAM

I have been a Transkribus user since 2017, a certified Transkribus trainer since 2020, and an elected director of READ-COOP SCE — the European cooperative behind the Transkribus platform — since 2023; from May 2025 I serve as Chair of the Board.

My work in this area is infrastructural in two senses simultaneously. Technically, it involves training HTR and ATR models, developing transcription and segmentation protocols for early modern handwriting, evaluating recognition engines against each other and against large-language-model-based approaches, and testing automatic subject indexing on normative corpora. The public ATR models I have coordinated — the Dutch Dean, Dutch Demeter, Dutchess, and the NeoLatin Ravenstein collections among others — are the practical output of this technical work.

Institutionally, the question is one I find at least as pressing: who owns and sustains the digital infrastructure on which humanities scholarship increasingly depends? READ-COOP’s cooperative model — a Societas Cooperativa Europaea with university, archive, and individual members — is my practical answer to that question. It is also an experiment in whether public-good digital infrastructure can survive the end of project-based EU funding, and whether the scholarly communities that depend on it can be persuaded to take collective ownership of it.

The HAICu work-package at the University of Twente applies these commitments to a specific regional corpus — the resolutions of the States of Overijssel (1528/1578–1795) — while HISMET addresses the archive-side metadata problem that typically falls outside the scope of researcher-oriented digitisation projects.

I write about these matters for Open Research Europe, the International Journal of Digital Humanities, JDMDH, and the Computational Humanities Research (CHR) proceedings, and produce regular Transkribus webinars in Dutch, English, and (in collaboration) French.

Related projects