Miracles, wonders and the art of coping with setbacks

Written by Anne Baillot

From September 1st to 5th, 2025, ATRIUM held its first Automatic Text Recognition Summer School at the Centre Marc Bloch in Berlin. With a focus on advanced projects, it attracted applicants from all over the world. In the selection process, projects were picked that faced similar challenges: text printed in columns, manuscripts displaying several layers of text, several handwritings, several languages, several alphabets. The trainers David Lassner and Floriane Chiffoleau developed training units that specifically addressed these issues.

The scientific richness of the participants’ data became clear from the first day on where their presentations mostly aimed at identifying recognition issues for the week to come. It was clear then that the lack of existing reusable models would be a major challenge for a range of projects. Interestingly, this challenge burdened not only lesser resourced scripts like Glagolithic or polytonic Greek, but also English - one can only assume the reasons vary depending on the script, and it is very likely that the lack of English models has more to do with access policies than with an actual lack of resources. 

Indeed, making the research output open for reuse was a key mantra of the Summer School and the reason why we favored using Kraken and eScriptorium as a platform for transcription. Discussions on making data and models available for reuse showed the cultural and infrastructural divide among European countries when it comes to implementing open science principles. 

After a few days, the main features of the platform had no secret left for the participants, whose expectations expectedly rose as they grew better at using it. Yet the week was mainly marked by highs and lows alongside efforts in model training and reassessments of the types of wonders or miracles that could be expected from automatic processes. Reconvening every morning, the first question asked in the room was whether the overnight training had been fruitful or not – a ritual we had to put an end to on the Friday as the summer school was over, but that was pursued by different participants in the aftermath.

The participants arrived with big hopes for their datasets. In their closing presentations on Friday, many reflected on the gap between their expectations and their actual gain. Small victories were all the more celebrated. All in all, the participants left better equipped to approach their goals realistically, with a tailored methodological basis, a group of like-minded and helpful trainers to turn to in their upcoming endeavours – and the wish to make their work as reusable as possible by the community. ATRIUM has gained great ambassadors of its values along the way!