ATRIUM Second Mutual Learning Exercise
By Carol Delmazo
June 9, Room T103, Troias building of the Athens University of Economics and Business (AUEB) in Athens, Greece.
Colourful post-its, an active whiteboard, checklists, a workflow divided into five modules running on the computers, lively discussions, and a great atmosphere. Collaboration mode: ON. The ATRIUM Second Mutual Learning Exercise (MLE) took place successfully, bringing together energy, teamwork, and diverse ideas on how to improve the Research Spotlight.
The purpose of the MLE is to promote mutual learning and the exchange of experiences and best practices between service and data providers, ensuring that improvements are driven by the contributions of both sides. This was made possible by a group of organisers from the ATRIUM project, who tailors the MLE format to specific needs of each service.
The Research Spotlight is an ontology-driven, automated workflow that transforms scholarly publications into structured knowledge graphs. Through a modular, notebook-based setup, it breaks down complex academic texts into entities and relations, such as methods, activities, and goals, following Semantic Web linked data standards. It serves as both a data processing pipeline and a pedagogical tool for teaching text analysis, knowledge organisation, and semantic technologies.
The service allows users to experiment with each module of the workflow, observe intermediate results, and gain hands-on insight into how scholarly knowledge can be structured and connected in a systematic, meaningful way. And this was exactly what the data providers did during the event. But that wasn’t all!
Bringing together a diverse group of participants from institutions such as AUEB , ATHENA Research Center (service providers), EKT , ADS and IBL PAN (data providers), alongside OPERAS , IBL PAN, Net7 and Foxcub (ATRIUM Organisers), the one-day event was a true opportunity for everyone to learn from one another in an interactive and participatory environment. It was hosted and co-organised by the UNESCO Chair on Digital Methods for the Humanities and Social Sciences, AUEB .
Breaking down the 2nd MLE
After an introduction to the ATRIUM project and a quick ice-breaker exercise, data providers delivered initial presentations focused on describing their institutions’ repositories and internal workflows, allowing participants to better understand their colleagues’ work realities. This was followed by an introduction to Scholarly Knowledge Graphs, setting essential context for the day’s activities.
Before diving into the Research Spotlight, Exercise 1: “Do I know the data I provide?” encouraged reflection on each institution’s data through a structured self-assessment checklist and a subsequent group discussion. Participants evaluated the quality and consistency of their (meta)data. They also analysed aspects such as open access/licensing, disciplinary classification, interoperability issues, and multilingual challenges, creating a collaborative space to share both difficulties and best practices.
Next up was Exercise 2: “The data journey in the Research Spotlight,” an immersive, hands-on session. Participants used Jupyter notebooks to run their datasets through the workflow’s modules, which included entity extraction, entity disambiguation, entity linking, relation extraction, and RDF generation. Using a colour-coded system of green, yellow, and pink sticky notes, the group collaboratively mapped out the workflow’s strengths, critical gaps, and inconsistencies. This allowed service providers to see exactly where the pipeline thrived and where adjustments are necessary.
The day concluded with Exercise 3: “The future of data provision in the Research Spotlight,” where the focus shifted to upcoming developments. Participants explored possible future capabilities, namely a hybrid Retrieval-Augmented Generation (RAG) implementation, the expansion of entity types, and domain-specific training datasets. Through a collective voting system, data providers mapped out their priorities and expectations, highlighting which future technical upgrades would best align with their institutions’ needs.
Why was the event successful?
For data providers, the MLE was a rare and valuable opportunity to step back and reflect deeply on the data they handle on a daily basis. Conversely, for service providers, it was a crucial chance to understand the subtle nuances of managing this data, learning firsthand how each step of the Research Spotlight workflow can be optimised to better serve the community.
Ultimately, the true success of the ATRIUM Second MLE was found not just in the code or the data, but in the collaboration itself. As noted in the foundational design of the event, mutual learning fundamentally relies on mutual listening. Sharing and learning works best when participants find a space to both offer their expertise and absorb new ideas. Leaving Athens, organisers, data and service providers carried home more than just technical insights; they left with a reinforced network, being reminded that the most complex technological challenges are best solved through shared expertise.