NLP Summer School 2024 report: Computational Language Technologies for Medievalists – A Week of discover something new
From July 8-12, 2024, the University of Graz hosted the Summer School on “Computational Language Technologies for Medievalists.” This intensive five-day program brought together a group of 22 young scholars, ranging from MA to PostDoc, from Canada to Turkey, and from novice programmers to experienced researchers. However, all united by their passion for medieval studies and their interest in harnessing the power of computational methods. You can find the list of their projects and posters in the main page of the school.
Our charming city provided a fitting but a bit too warm backdrop for this blend of medieval texts and cutting-edge technology. As participants gathered at the 3rd floor of the Zentrum für Informationsmodellierung, there was a palpable sense of excitement and anticipation. The program promised a journey through the of Natural Language Processing, Topic Modeling, Named Entity Recognition, Text Reuse, and Large Language Models – all tailored specifically for historians, especially for medievalists.
Day one kicked off with an introduction to Python programming, led by Lucija Krusic. For many, this was their first foray into coding, and the enthusiasm was evident. As one participant noted, “Today the most valuable takeaway is my first approach to Python.” The day progressed with sessions on NLP methods and corpus building, culminating in a captivating keynote by Jean-Baptiste Camps (Ecole des Cahrtes, Paris) on computational methods and medieval manuscripts, guiding us through his work in large scale and long range analysis of manuscripts and text of the chivalric romance in France. Despite the challenging heat, participants left invigorated.
The second day delved into the world of Topic Modeling, guided by Gabriel Viehhauser (University of Vienna) and Martina Scholger (University of Graz). Participants grappled with concepts like Latent Dirichlet Allocation and BERTocip, applying these techniques to real datasets. The hands-on sessions were particularly appreciated, with one attendee remarking, “Tools to work with our own data in the future.” The day’s lessons sparked creativity, with another participant noting, “BERT might better work for medieval corpora than word2vec.”
Day three brought the fascinating world of Named Entity Recognition to the forefront. Ismail Prada (University of Bern), supported by Selina Galka and my DiDip team member Sandy Aoun (both University of Graz) led engaging sessions that deepended our understanding of what named entities really are in medieval studies (yes, nesting is part of it!) and had participants working with NLP libraries like NLTK, spaCy and, in particular, flair, and annotation tools like Inception. The practical applications of NER in medieval studies became clear, with one scholar enthusing, “NER has great applicability to my work.” The day concluded with a lively poster session and a delightful dinner, fostering connections among the participants.
Text Reuse and Stylometry took center stage on the fourth day, led by Jeroen de Gussem (University of Gent) and me (i.e. Tamás Kovács, University of Graz) while my DiDip colleague Anguelos Nicolaou (University of Graz) explained the mathematical background of each pocesses. Participants explored algorithms for detecting text borrowing and attributing authorship, applying these methods to medieval texts. The complexity of the topic was evident, but so was the excitement. As one attendee put it, “That for Topic Modeling you are interested in the words with lots of meaning and for stylometry the filler words might be what defines the style.” The day concluded with city tour of Graz, allowing participants to appreciate the historical context of their studies.
The final day focused on the future of computational linguistics, with Axel Pichler (University of Stuttgart) and the DiDip PI Georg Vogeler (University of Graz) discussing Large Language Models. This forward-looking session sparked intense discussions about the potential impact of AI on medieval studies. In the afternoon, we went to a nearby Buschenschank:
Despite the heavy rain after the Buschenschank, spirits remained high as participants reflected on their week-long journey. One summed it up beautifully: “Met interesting people, learnt some things, got a couple of ideas for research. Felt really welcome; you are great hosts.” Throughout the week, we had remain flexible, adjusting to participants’ needs and the challenging weather conditions. We hoped, that the mix of lectures, hands-on sessions, and social activities created a rich learning environment that went beyond mere technical skills. We might have succeded with it, as one participant noted, “I learned at least as much there as I did in class and I felt very welcome.”
The Summer School’s academic richness was further enhanced by three themed poster sessions, each offering a window into the diverse applications of computational methods in medieval studies over a bottle of beer. Tuesday’s session, “Social Networks, Power Dynamics, and Institutions,” featured fascinating projects ranging from Sigismund of Tyrol’s court practices to social networks in 12th-century Byzantium. These posters demonstrated how digital tools can illuminate complex social and political structures of the medieval world. Wednesday’s session, “Historical Data Extraction and Analysis,” showcased innovative approaches to extracting meaning from historical texts. Presentations on topics such as mathematics vocabulary in Middle French and the impact of plague on land tenancy in St. Pölten highlighted the power of computational methods in uncovering hidden patterns in historical data. Thursday’s session, “Text Reuse and Language Evolution,” explored the intricate ways texts were adapted and language evolved across medieval cultures. From analyzing the documentary language of the Reformpäpste to problematizing vector semantics in medieval love concepts, these posters pushed the boundaries of traditional textual analysis. The poster sessions not only provided a platform for participants to present their work but also sparked lively discussions and collaborations, enriching the overall learning experience. You can find the list of their projects and posters in the main page of the school.
As the Summer School concluded, we left the participants with newfound skills, fresh perspectives, and a network of colleagues from across the field. The extensive notebooks and materials provided ensure that the learning will continue long after the participants have returned home. You can find it at Tamás’s GitHub repo. This week in Graz not only bridged the gap between medieval studies and modern computational methods but also fostered a community of scholars ready to push the boundaries of their field.
The “Computational Language Technologies for Medievalists” Summer School demonstrates the growing interest in digital humanities and the potential for computational methods to unlock new insights in medieval studies. As we look to the future, we hope that events like this will play a crucial role in shaping the next generation of medieval scholars, equipping them with the tools to explore medieval texts in innovative ways.
Thanks to the support of the Postdoctoral Office of the University of Graz (https://postdoc.uni-graz.at/de/) for the Event funding. Thanks to the support of the university network HFDT we could offer 10 selected applicants financial support for accommodation in Graz.
OpenEdition suggests that you cite this post as follows:
ktamas (August 5, 2024). NLP Summer School 2024 report: Computational Language Technologies for Medievalists – A Week of discover something new. DiDip. Retrieved November 11, 2024 from https://didip.hypotheses.org/2780