Skip to main content

ACDH Lecture 12.1

From Punch Cards to Prompt Engineering: The MHDBDB and the Future of Semantic Annotation with LLMs

with Katharina Zeppezauer-Wachauer , Senior Scientist at the University of Salzburg and Julia Hintersteiner , Senior Scientist and PhD candidate at the University of Salzburg

When: Tuesday, January 20th 2026; 16:45 - 18:15

Where: University of Vienna, Universitätsring 1, 1010 Vienna,
Hörsaal 6 Franz König Hauptgebäude, Tiefparterre, Stiege 9 Hof 5 and online (u:stream).

Registration: closed.


The Mittelhochdeutsche Begriffsdatenbank ( MHDBDB ) has supported generations of scholars working with medieval German texts. After decades of development through relational databases and, more recently, RDF-based graph architectures, the MHDBDB has undergone a complete technological redesign – now centering on a TEI-first model that directly supports LLM-based research workflows.

This talk presents the rationale and lessons learned behind this shift: the increasing importance of structured, AI-readable data; the technical and infrastructural limitations of high-complexity standoff models; and the sobering reality that managing billions of RDF triples requires levels of personnel and funding rarely available to humanities projects. In this context, AI – particularly Large Language Models – emerges not only as a supplementary tool, but as a game changer. With LLMs now playing a growing role in annotation, search, and interpretation, classical TEI-XML has proven to be the ideal environment for Expert-in-the-Loop workflows: it enables transparent, semantically rich, and interoperable data structures that are both human-legible and machine-actionable – bridging philological precision and AI scalability.

Literature:

  • Alan van Beek. 2025. Von historischen Daten zu KI – Bericht zum interdisziplinären Workshop. Digital Humanities an der Universität Salzburg. DOI: https://doi.org/10.58079/155ob . Darin verlinkt: Foliensatz und Aufzeichnung des Vortrags von Christopher Pollin und Katharina Zeppezauer-Wachauer (hinterer Teil zur technischen Neuausrichtung der MHDBDB).
  • Alan van Beek und Julia Hintersteiner. 2025. “Poc, the Black Knight” – Wie voreingenommen sind LLMs bei mittelalterlicher Literatur? Digital Humanities an der Universität Salzburg. DOI: https://doi.org/10.58079/153my
  • Julia Hintersteiner, Michael Göggelmann, Alan van Beek, Katharina Zeppezauer-Wachauer. 2024. Von der Datenpflege zur Schnittstellenentwicklung: Eine Evaluation des Datenkolosses MHDBDB. Digital Humanities an der Universität Salzburg. DOI:  https://doi.org/10.58079/12p9n
  • Katharina Zeppezauer-Wachauer. 2022. “50 Jahre Mittelhochdeutsche Begriffsdatenbank (MHDBDB). Eine Jubiläums-Zeitreise zwischen Lochkarten, Pixel-Drachen, Relationaler Datenbank und Graphdaten.” In: Beiträge zur mediävistischen Erzählforschung, Digitale Mediävistik. Perspektiven der Digital Humanities für die Altgermanistik, no. Themenheft 12 (2022): 161–186. DOI: https://doi.org/10.25619/BmE20223203 .

Please RSVP for the onsite-lecture / register for the online lecture.

This ACDH Lecture is organized in close collaboration with the University of Vienna within the framework of the Digital Humanities Lecture Circuit .

about the speakers

Katharina Zeppezauer-Wachauer is a Senior Scientist at the University of Salzburg, where she has been part of the Mittelhochdeutsche Begriffsdatenbank (MHDBDB) team since 2010 and has coordinated the project since 2016.

Her research focuses on corpus linguistics, computational text semantics, and the application of AI and LLMs to historical corpora, with an emphasis on semantic modeling, narrative structures, and FAIR data practices.

She holds a PhD in German Studies from the University of Graz with a specialization in Digital Humanities and serves as the University of Salzburg’s spokesperson for the Austrian infrastructure consortium CLARIAH-AT.

Learn more about Katharina Zeppezauer-Wachauer:

Julia Hintersteiner, MA BA, is a Senior Scientist and PhD candidate at the University of Salzburg.

Working at the interface of Medieval Studies and Digital Humanities, she is dedicated to the advancement of the Middle High German Conceptual Database (MHDBDB). Her doctoral thesis directly addresses this resource, focusing on word sense disambiguation and semantic annotation to refine the database’s capabilities.

Julia’s expertise extends to the Wenceslas Bible , where she works on the automated indexing of text-image relations and iconographic annotation. She holds an MA from the University of Graz (2023), having also studied in Salzburg and Utrecht.

Learn more about Julia Hintersteiner: