My research combines qualitative and quantitative methods to analyze the relationship between linguistic and extralinguistic patterns in texts and dialogues. In interdisciplinary projects, I investigate how corpus-linguistic, experimental, and computational approaches advance the study of how humans and machines produce and interpret patterns of language use.

Name-based word formation patterns in German

During my work as a postdoctoral researcher in the research unit Project “Musterhaftigkeit – Linguistic Creativity and Variation” (funded by the Rhineland-Palatinate Research Initiative, 2019–2023), I developed an interest in the role of individual and collective extralinguistic knowledge in the production and interpretation of word-formation patterns. As part of my habilitation project, I examined how different contexts, such as text types and media (e.g., newspapers vs. social media) or discourse domains (politics, sports, and celebrities), affect the production and interpretation of name-based compounds and compound-like patterns in German. The empirical focus is, first, on a type of compound in German that has received little attention to date, featuring a personal name as the second constituent (e.g., Kabinen-Merkel), and second, on name blends as a compound-like phenomenon, such as Merkozy from (Angela) Merkel and (Nicolas) Sarkozy. Methodologically, the study combines corpus-linguistic and experimental approaches.

Patterns of interaction with language models

This project builds on my habilitation and PhD theses, extending their methods and research questions to human-machine interaction. Current work examines the capability of language models to interpret common ground phenomena and domain-specific meanings across various discourse domains, including both text corpora and naturally occurring dialogues. A practical strand asks how users’ trust can be calibrated depending on the models’ demonstrated performance at producing and identifying context-sensitive linguistic patterns (see e.g., Belosevic 2025, Belosevic & Buschmeier 2024).

The linguistic construction of trust, mistrust and fear in German media

As part of my PhD project, I examined how trust, mistrust, and fear were constructed in German media coverage of the refugee debate (2015–2017). Drawing on the concept of argumentation patterns (topoi), I operationalized the sociological concept of trust for a discourse-analytic study. Using corpus-linguistic methods, I tested whether trust toward refugees changed from 2015 to 2017 across argumentation patterns identified through manual annotation. Subsequently, I applied this approach to other domains of public discourse, including Euroscepticism and the public debate about COVID-19. I also co-organized the conference on Language and Trust and co-edited the volume Sprache und Vertrauen. Facetten der linguistischen und interdisziplinären Vertrauensforschung.