Dr. Tejaswini Deoskar

Associate Professor

t.deoskar@uu.nl

My research focuses on Natural Language Processing, which involves creating technologies that understand and work with human language, and Computational Linguistics, which applies computational methods to study language.  I have a particular focus on syntax, the grammatical rules that govern the structure of sentences, as well as other structural properties of language. This includes grammar induction, which aims to learn hidden grammatical structures that underlie language from data, and examining how these structures relate to meaning (known as the syntax-semantics interface). In my research, I employ methods from computer science, such as machine learning and neural networks, along with insights from linguistics and cognitive science. I am particularly interested in tackling rare and complex language patterns that current NLP models often struggle with, known as “long-tail” phenomena. Amongst other topics, my current research interests span multi-modal models that use images and videos (e.g. how can we “ground” the learning of grammar by using images and video data in addition to language), and fairness and regulation of NLP systems and Large Language Models as they become more widely used, using techniques like watermarking and fingerprinting of models to help with this.