Leon Derczynski

Leon Derczynski is a researcher at the IT University of Copenhagen with a focus on Natural Language Processing and computational linguistics. His work spans multiple NLP subfields including temporal information extraction, misinformation detection, and social media analysis. He has contributed to the development of NLP resources for Danish and Nordic languages, and created frameworks like garak for model security probing.

Research interests include:

• Temporal relation classification and time expression modeling
• Social media analysis and misinformation detection
• Model efficiency and resource-aware NLP
• Scandinavian language processing
• Ethical considerations in NLP

Publications highlight trends in transformer architecture optimization, set-to-sequence modeling, and abusive language detection. His work frequently appears in top venues like Transactions of the Association for Computational Linguistics, EMNLP, and COLING.

Key collaborations include work with Kalina Bontcheva on rumor evaluation, Manuel R. Ciosici on efficient NLP methods, and Erick Galinkin on model security. He has also contributed to datasets like the Danish Gigaword Corpus and evaluation frameworks like Risk Cards for model deployment assessment.