What is it about?
LeSum is a system designed to summarize long and complex Indian legal judgments. It works in two steps: first, it picks out the most important sentences using a smart graph‑based method that breaks down lengthy sentences into smaller, coherent parts and selects those that convey meaningful information preserving the overall context while avoiding repetition. Then, it uses large language models (LLMs) to rewrite these into structured summaries that highlight key elements of a case—such as facts, statutes, arguments, precedents, and the final decision. This approach reduces the cost and effort of processing judgments while making legal information more accessible, especially in settings where annotated legal data is limited.
Featured Image
Photo by Aleix Ventayol on Unsplash
Why is it important?
LeSum helps make long and complex Indian legal judgments easier to process, understand, and use. Legal documents are often lengthy, filled with technical language, and time‑consuming to read. By combining extractive and abstractive summarization, LeSum reduces the length of the text by over 91%. Whereas, the extractive summarization module itself compresses text by nearly 60%, lowering both computational cost and effort for subsequent abstractive summarization. It produces clear, structured summaries that highlight the most critical parts of a case such as facts, statutes, arguments, precedents, and decisions. This improves the overall readability and informativeness of the summaries, making them more accessible to lawyers, researchers, students, and even the general public. Importantly, its zero‑shot learning capability means it can work effectively even in low‑resource settings where annotated legal data is unavailable, offering a scalable solution for improving access to justice and legal knowledge.
Perspectives
LeSum addresses a genuine real‑world challenge in the legal domain by helping stakeholders such as lawyers, judges, students, and researchers scan voluminous judgments filled with dense and convoluted sentences. It generates structured summaries that capture essential components such as case metadata, facts, statutes, arguments, precedents, and decisions. Through extensive experiments involving comparisons with related works on multiple metrics, as well as human evaluations, LeSum demonstrates its potential to democratize access to legal data and support justice systems. Importantly, it offers a cost‑effective solution well‑suited for low‑resource scenarios where annotated legal data is scarce.
Wazib Ansar
Read the Original
This page is a summary of: LeSum: Cost-effective LLM-driven hybrid summarization of Indian legal judgments, Expert Systems with Applications, January 2027, Elsevier,
DOI: 10.1016/j.eswa.2026.133683.
You can read the full text:
Contributors
The following have contributed to this page
