From Text to Graph: Leveraging Graph Neural Networks for Explainability in NLP