Conference Agenda
Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
|
Session Overview |
| Session | ||
Freitag 1:5: Freitag 1:5 – Automatisierung und KI
| ||
    | ||
LLM-Assisted Metadata Extraction and Normalization for Historical Correspondence: A Multi-Stage Pipeline Approach Universität Graz, Österreich This paper presents an LLM-assisted approach to automating metadata extraction from historical correspondence. The methodology was developed using the multilingual Joseph von Hammer-Purgstall correspondence collection (1774-1856), which presents typical challenges of historical documents including orthographic variations, archaic date formats, and contextual ambiguities. Rather than relying solely on LLMs, the proposed multi-stage pipeline combines automated processing with established digital humanities standards. The system extracts core metadata elements (sender, addressee, date, location), links entities to authoritative databases like GND and GeoNames, and uses contextual analysis to resolve ambiguities based on historical context. The final output conforms to the Correspondence Metadata Interchange Format (CMIF) for integration with existing digital humanities platforms like correspSearch. Automatic Annotation and Modelling of Works in Eighteenth-Century (Music) Theatre Karl-Franzens-Universität Graz, Österreich This paper presents a hybrid approach to annotating and modeling references to theatrical works in eighteenth-century theatre chronicles. As part of the FWF-funded project GuDiE (2024–2028), a digital critical edition of the Gumpenhuber chronicles is being developed using TEI/XML, alongside an RDF database based on performing arts ontologies. Due to the fluid and hybrid nature of historical theatre practices—such as pasticcio or cross-genre adaptations—works are modeled not as abstract entities but as expressions enriched with content types (e.g., music, text, choreography). To support the annotation process, Large Language Models (LLMs) were used for semi-automatic Named Entity Recognition and linking to a curated authority register. Results show that LLMs can reduce manual effort, though challenges such as ambiguity, disambiguation, and prompt sensitivity remain. This contribution reflects on the methodological and ontological implications of combining structured data modeling with historical complexity in digital scholarly editing. Keyness Measures und BERTopic kombiniert: Eine Distinktivitätsanalyse von Subgenres des französischen Romans Universität Trier, Germany Der vorliegende Beitrag stellt einen hybriden Analyseansatz vor, der word-embedding-gestützte Topic-Modeling-Verfahren mit Distinktivitätsmaßen kombiniert, um Textgruppen zu vergleichen. Wir haben ein Topic-Modell für eine Sammlung von 600 französischen Romanen trainiert, distinktive Topics für jedes Subgenre ermittelt und diese Topics dann mit manuell erstellten textuellen Gattungsprofilen verglichen, um die Leistung verschiedener Distinktivitätsmaße bei der Identifizierung von distinktiven Topics zu evaluieren. | ||