Project - Methods for Causality in NLP
The primary objective of scientific research is to comprehend causal relationships. Although causality plays a crucial role in life and social sciences, it has not received the same level of significance in Natural Language Processing (NLP), where the focus has traditionally been more on predictive tasks. This distinction is now starting to diminish, giving rise to an interdisciplinary research field that converges causal inference with language processing. Nevertheless, investigations into causality within NLP are currently dispersed across various domains, lacking unified definitions, benchmark datasets, and clear articulations of challenges and opportunities in applying causal inference to the textual domain, which possesses unique properties. In this research project, we will explore ideas to apply causal mechanisms to Large Language Models to enhance their performance, making them more robust and transparent progressing in the direction of making AI more trustworthy.
In progress