Abstractive Summarization of Indian Legal Judgments

Abstract

This paper reports the effectiveness of a method using T5 to generate an abstractive summarisation of Indian legal judgments. When a legally qualified person manually performs summarisation, the created summary always depends on the person's expertise in performing the task. So an automatic legal summarization is used to perform this task more accurately. The system generates an abstractive summary of 350 judgments taken from the Honorable Supreme Court of India. The dataset has been created manually with a lawyer's help, and the result evaluation is performed using the ROUGE score, which gave a precision of 0.54955 for Rouge-L.