Abstract
Indian court judgment reports frequently include complicated words and sentences, making it difficult for the general public and legal experts to understand these legal documents. Legal organizations hire legal experts to provide summaries of complex and lengthy legal texts. Hence, a variety of techniques have been created to construct the summaries. In this research, we utilized the InLegalBERT model, originally trained to perform Legal Statute Identification, Semantic Segmentation, and Court Judgment Prediction tasks on Indian legal documents. In addition to these three tasks, the main goal of this research is to suggest a novel approach to use InLegalBERT to perform downstream tasks of summarization. To evaluate the effectiveness of our summarization strategy, we employed four different models: Legal Pegasus, T5 base, BART, and BERT. Based on the ROUGE-L F1 scores, the suggested approach using the InLegalBERT model is performing the best for Indian legal document summarization with a precision of 0.3022 and a recall of 0.664. Evaluation result Rouge1 F1 is 0.4226, Rouge2 F1 is 0.2604 and RougeL F1 is 0.4023