LINGUISTICS-INFORMED NATURAL LANGUAGE PROCESSING MODELS

Abstract

Natural Language Processing (NLP) has witnessed remarkable advancements in recent years, owing much of its progress to deep learning models such as transformers. These models have demonstrated their exceptional capability in understanding and generating human language, but their performance can be further improved by incorporating linguistic insights, specifically constituency information. This thesis presents a novel approach to enhance extractive QA tasks by integrating constituency-based linguistic knowledge with state-of-the-art Heterogeneous Graph Transformer (HGT) models.