Advancing Legal Document Summarization: Introducing an Approach Using a Recursive Summarization Algorithm

Researchers are increasingly focusing on legal text processing due to the abundance of legal information online. Understanding lengthy legal documents has become challenging due to the sheer volume of data. Therefore, there is a growing need for systems that can help legal professionals and the general public to access relevant legal information easily. Additionally, studying the integration of LLMs and dealing with the length of the document in this field could be a crucial aspect of research. To bridge this gap, we propose a novel framework, RecSumm, which leverages a recursive summarization algorithm to address the challenge of summarizing long legal documents. we evaluated the performance of RecSumm by comparing it with LLMs and legal domain-specific models using ROUGE and BLANC scores. Our evaluation reveals that RecSumm outperforms LLMs like ChatGPT and Gemini and legal domain-specific models like Legal Pegasus and Legal-LED in preserving crucial information while generating concise summaries. Furthermore, this study explores the limitations of traditional ROUGE scores and investigates the potential integration of LLMs within the summarization process. These insights pave the way for future research to advance text summarization techniques specifically for the legal domain.