Research
Four themes connect most of our work. Many projects touch more than one.
Health and clinical text
Language technology for hospitals and health research: clinical notes, drug safety and patient privacy.
Much of what is known about patients is written as free text: discharge letters, clinical notes, reports of side effects. We build methods that help clinicians and researchers use this text, while protecting patient privacy and working in the languages spoken in Swiss hospitals.
Recent work includes drafting discharge summaries in Italian together with clinicians, finding adverse drug reactions to support drug safety, and removing personal information from clinical records.
Languages, dialects and multilinguality
How language models deal with dialects, scripts and less-resourced languages, and how to make them fairer across languages.
Language models work best for a handful of large languages. We study why, starting from the text itself: how words are split into pieces (tokenization), how scripts and dialects are represented inside models, and how much information text carries in different languages.
The goal is language technology that serves speakers of all languages and dialects more equally. Our work on tokenization bias in dialects received a Senior Area Chair award at EMNLP 2025.
Understanding and trusting language models
When language models get things wrong, and how to tell: hallucinations, machine-generated text, uncertainty.
Language models are now used to search, summarise and write, but they can be confidently wrong. We study how they behave and where their limits are: spotting invented facts (hallucinations), telling machine-generated text from human writing, testing whether models understand cause and effect, and checking systems that answer questions from documents.
We also look inside models to understand how they reach their outputs, for example at which point a model decides which writing system to use.
Knowledge from scientific literature
Turning scientific papers into structured, searchable knowledge, and making science easier to read.
Science is published faster than anyone can read it. We build tools that find facts in papers, such as genes, diseases or traits of species, and turn them into data that researchers can search and reuse. A long-running collaboration with biologists builds a database of arthropod traits from the literature.
We also work on simplifying and translating scientific text for wider audiences, and on visual ways to explore large collections of publications.