معرفی
Jannis Vamvas is a Lecturer and active researcher in the Department of Computational Linguistics at the University of Zurich, Switzerland. With a PhD thesis titled Model-Based Evaluation of Multilinguality, he specializes in deep learning for natural language processing, with particular emphasis on multilingual language models and their evaluation.
- Academic Roles
- Current Lecturer for Large Language Models, Text Generation, and Mathematical Foundations at UZH
- Postdoctoral Researcher (MUTAMUR Project, 2023)
- PhD in Computational Linguistics (2019-2023, supervised by Rico Sennrich, Lena A. Jäger, and Martin Volk)
- Education
- M.Sc. in Computational Linguistics (major) and Computer Science (minor) at LMU Munich (2017-2019)
- B.A. in Computer Science and Philosophy from University of Basel (2011-2015)
His research focuses on multilingual language models, particularly SwissBERT for Swiss languages, and methods for evaluating translation quality through approaches like Contrastive Conditioning and Translation Cross-Likelihood. He explores how to adapt models to local data, detect translation errors, and improve document translation through source-primed conversation techniques.
Recent publications highlight his work on:
- Translation direction detection
- Hallucination mitigation in machine translation
- Minimum Bayes Risk decoding
- Swiss multilingual language modeling
- Semantic difference recognition
Scientific recognition includes:
- EAMT 2024 Highly Commended Thesis
- Best Paper Award at SwissText 2025
- Best Scientific Paper Award at SwissText 2024
He teaches advanced courses on Large Language Models, Text Generation, and Mathematical Foundations, and has served as organizer of the Computational Linguistics Colloquium at UZH. Current projects include the InvestigaDiff initiative and machine translation for Romansh idioms.
