Project Profile Using LLMs to Retrieve and Analyse Speech Acts

The project aims to conduct a research seminar titled “Using LLMs to Retrieve and Analyse Speech Acts” in the Winter Semester 2026/27. The seminar develops core competencies in linguistic research and provides knowledge of AI in general, as well as a comparative understanding of specific Large Language Models (LLMs). At the same time, it contributes directly to ongoing research on the use of AI in linguistics. The project places particular emphasis on the close integration of research and teaching.

Within the seminar, students train and evaluate different LLMs for the identification of speech acts. LLMs offer the potential to fundamentally extend corpus pragmatics. Unlike traditional search methods, they can identify not only identical patterns but also similar ones—without requiring that similarity be explicitly defined in advance. This potential, however, remains underutilized. The use of LLMs in corpus pragmatics is still in an early stage of development, even though initial studies point to promising directions.

Through in-depth engagement and independent research with LLMs in the seminar, students gain not only practical skills but also a deeper conceptual understanding of these systems. The teaching concept can be transferred to other areas in which the use of AI in research is currently being explored.