A strong showing for the University of Würzburg: it is involved in three out of ten research groups whose establishment or renewal has now been approved by the German Research Foundation.
A success for researchers in the humanities, social sciences, natural sciences and computer science at JMU: the German Research Foundation (DFG) has approved their applications for new research groups or extended the funding for existing ones. In total, the DFG is establishing five new research groups and providing them with funding totalling 27 million euros; five further groups have had their funding extended.
Research groups enable researchers to address current and pressing issues in their fields and to establish innovative lines of research. They receive funding for up to eight years. The DFG currently funds 186 research groups, seven clinical research groups and 16 ‘Kolleg’ research groups.
Semantic shifts in low-resource domains
A new research group, ‘Semantic shifts in low-resource domains’ (SILD), is being established at the University of Würzburg within the CAIDAS (Centre for Artificial Intelligence and Data Science) and ZPD (Centre for Philology and Digitality). The group’s co-chairs are Professors Fotis Jannidis, Chair of Digital Humanities and Modern German Literature at the University of Würzburg, and Andreas Hotho, Chair of Computer Science X (Data Science).
What this interdisciplinary group is investigating: How does language change over the course of history – in historical texts, literary works and academic discourses? Can these changes be automatically detected using AI? The team is developing new language models (Large Language Models) to identify shifts in linguistic meaning in collaboration with the humanities disciplines, where only limited amounts of data are available, thereby enabling new insights within the field.
A key focus is on the shift in meaning of expressions consisting of one or more words that together form a distinct meaning. The AI must learn to recognise these ‘word puzzles’ within the context of contemporary texts. To this end, the project utilises state-of-the-art AI methods from large language models, such as the Transformer architecture and knowledge graphs, as well as transfer learning methods – and, last but not least, innovative forms of visualising this data.
Three humanities research teams are working within SILD to investigate linguistic change:
- Using 800 German-language plays from the period between 1650 and 1830, the study examines changes in moral language;
- using Latin translations from Arabic, the project is investigating how Medieval Latin changed;
- using newspapers from Hong Kong, the development of a specific form of English in that region is being analysed.
In addition to the teams led by the two principal investigators, the following are also involved in the group:
- Professor Carolin Biewer, English Linguistics, University of Würzburg,
- Dr Katrin Dennerlein, German Literature, University of Würzburg,
- Professor Stephanie Evert, Computational Linguistics, University of Erlangen-Nuremberg,
- Professor Goran Glavaš, Natural Language Processing, University of Würzburg
- Professor Dag-Nikolaus Hasse, History of Philosophy, University of Würzburg
- Professor Daniel Keim, Visual Analytics, University of Konstanz.
