Legal Lay Summarization: Exploring Techniques and Introducing the LegalEase Dataset

This thesis investigates advancements in Natural Language Processing (NLP) for legal lay summarization by systematically analyzing the methodologies, datasets, and research findings presented in existing publications. The study reviews the current literature in the field, highlighting challenges in legal summarization, such as data scarcity and the complexity of legal language. A key contribution of this work is the creation of a specialized dataset, \textbf{LegalEase}, which is specifically designed to enhance model training for summarizing legal documents in layman’s terms. Key findings demonstrate that subdomain-specific datasets within the legal domain outperform more general legal domain datasets in enhancing the performance of NLP models in generating accurate and comprehensible legal summaries. The results presented in this thesis offer valuable insights and methodologies for future advancements in legal lay summarization.