Legal professionals routinely analyze lengthy and complex documents such as court judgments, contracts, petitions, and legal notices. Manual review of these documents is both time-consuming and susceptible to human error, making it difficult to efficiently identify critical legal information. Although recent advances in Natural Language Processing (NLP) have enabled automated text summarization, many existing solutions focus only on summary generation and provide limited support for comprehensive legal document analysis. This paper presents Briefly, an AI-powered legal document summarization and analysis system that combines transformer-based abstractive summarization with multiple NLP techniques to improve the accessibility of legal information. The proposed system is developed using a React-based frontend and a FastAPI backend, providing a scalable web application for processing legal documents in PDF, DOCX, TXT, and Markdown formats. The summarization module employs the pre-trained DistilBART transformer model from Hugging Face to generate concise summaries while maintaining computational efficiency. To overcome transformer input-length limitations, the system adopts a hierarchical two-pass summarization strategy consisting of sentence chunking, chunk importance scoring, batch summarization, and final summary compression. Beyond document summarization, the system integrates Named Entity Recognition (NER), rule-based Rhetorical Role Labeling (RRL), case brief generation, legal insight extraction, timeline generation, analytics, document-based question answering, and multi-format export functionality into a unified platform. The implementation further incorporates caching mechanisms and automatic CPU/GPU fallback to improve system responsiveness and reliability during document processing. The developed prototype successfully performs end-to-end legal document analysis through an intuitive web interface, demonstrating the practical integration of modern NLP techniques for legal information retrieval and summarization. The proposed system provides an efficient and scalable solution that can assist legal professionals, researchers, and students in understanding lengthy legal documents while reducing manual effort.
Raj et al. (Sun,) studied this question.