Search

The surge of digital legal documents has significantly expanded their usage. This has resulted in the sheer number of papers being used by various members of the judiciary, as well as by advocates and judicial officers. It can be incredibly challenging to keep up with all of them. Over four crore cases are still pending in Indian courts, and manually reviewing them can be a tedious and time-consuming process\cite{r1}. As machine learning has advanced, various text summarization models have been created to help legal professionals manage their documents. Due to the lack of publicly accessible datasets, it is difficult to fine-tune domain-independent models for Indian legal systems. The methodology proposed in this paper seeks to improve the overall performance of these models, and it also explores Indian legal documents' summarization techniques. In addition, this research also provides a study of the several summarization methods in-depth that have been on Indian legal documents, including PEGASUS, Bidirectional Auto- Regressive transformers (BART), TextRank, and Bidirectional Encoder Representations from Transformers (BERT). Through the process of extractive and abstract summarization, BART and PEGASUS will be able to gain a deeper understanding of the text normalization process. The outcomes of the text normalization process are evaluated by experts using the ROUGE metrics and multiple parameters. It shows that the proposed approach can work well in legal texts that have domain-independent frameworks. [ABSTRACT FROM AUTHOR]