Jan Pfister

Jan Pfister is a researcher at the Chair of Data Science (Informatik X), affiliated with the Faculty of Mathematics and Computer Science at the University of Würzburg. Holding an M.Sc. in Computer Science from 2021, he contributes to the DMIR group while maintaining the BibSonomy platform. His work bridges Natural Language Processing and Deep Learning, with a focus on structured sentiment analysis and German language modeling.

His research explores the integration of Large Language Models with Pointer Networks to extract fine-grained sentiment information. He has developed ModernGBERT, a German-only encoder model, and LLäMmlein, advancing NLP capabilities for the language. Pfister's recent publications highlight trends in encoder-decoder architectures, model-agnostic hallucination detection, and multi-label classification frameworks that refine LLM outputs.

He has taught courses including Information Retrieval (SS '22), Text Mining (WS '23), and Selected Topics in Machine Learning (SS'21, WS '21, WS'24). His collaborative projects span medical informatics, human-AI interaction, and social media analysis, often involving interdisciplinary teams and real-world data.