Extractive Summarization of Indian Legal Judgments: Bridging NLP and Generative AI for Socially Responsible Content Generation

In recent years, the fields of Natural Language Processing (NLP) and Generative Artificial Intelligence (AI) have seen significant advancements, particularly in the domain of legal text analysis and summarisation. This work explores the application of NLP and Generative AI techniques in the context of Indian legal judgments, focusing on extractive summarisation methods and their implications for socially responsible content generation. This work uses state-of-the-art NLP models, including T5 and GPT-2, to generate extractive summaries of Indian legal judgments. These models are fine-tuned on the manually created extractive summaries to enhance their performance in summarizing legal content. The proposed approach is evaluated using standard metrics such as ROUGE and BERTScore. Results demonstrate the method’s effectiveness, with significant improvements observed over baseline models. The approach achieves scores of 0.34 for ROUGE-1, 0.15 for ROUGE-2, and 0.31 for ROUGE-L metrics, along with 0.26 for PBERT, 0.41 for RBERT, and 0.31 for FBERT for the T5 model. For GPT-2, the corresponding values are 0.27 for ROUGE-1, 0.12 for ROUGE-2, and 0.24 for ROUGE-L metrics, along with 0.33 for PBERT, 0.25 for RBERT, and 0.28 for FBERT. This study contributes to the advancement of extractive summarisation techniques for Indian legal judgments, providing a robust framework for generating accurate and concise summaries that capture the essence of complex legal documents by leveraging advancements in NLP and Generative AI to generate socially responsible summaries.