Large Language Model based Legal Text Summar...

Large Language Model based Legal Text Summarization (LLTS)

Author Name : Megha Parate

ABSTRACT Legal judgments are generally very long, and relevant information is often scattered throughout the text. Summarizing a legal judgment requires capturing crucial details comprehensively from the lengthy content. Abstractive-summarization models based on pre-trained language often encounter limitations in handling extended input texts. Furthermore, these models struggle to seamlessly integrate technical terms and specific topics prevalent in legal judgments. In this paper, a new dataset, recently developed and trained on Indian supreme court case documents, was utilized to maintain relevance and capture legal nuances in the summary. OpenAI's fine-tuned GPT-3.5-turbo model, along with a map-reduce framework, was employed to overcome token limitations. The experimental results revealed a remarkable improvement in rouge scores compared to existing models.