Research theme · part of Robotics and Semantic Systems

Multimodal AI and Privacy

Parsing, language models and language as a robot interface, and what happens to text when personal data must be removed from it.

NLPMulti-Agent SystemsData PrivacyMultimodal AIMedia & Language

Natural language processing in the broad sense: parsing, semantic role labelling, word embeddings and language models, plus the use of language as an interface to robots: the group’s oldest line of work and still one of its largest. The current focus is privacy: automatic pseudonymisation of freely written Swedish text, so that research data containing personal information can be opened up for reuse, and studying what that rewriting does to a text’s readability and research value.

“Language is often the easiest way to talk to a robot, but it's also one of the easiest ways to leak someone's identity. We work on both sides of that problem at once.”
Xuan-Son (Sonny) Vu, Asst. Prof.Associate senior lecturer