LEGAL DOCUMENT CLASSIFICATION

Legal professionals often manage large volumes of unstructured documents such as judgments, contracts, and legal notices, which require accurate classification for legal research, compliance, and judicial processes. Manual classification is tedious, time-consuming, and prone to inconsistencies. This project, titled “Legal Document Analysis and Classification Using NLP and Deep Learning”, proposes an automated solution utilizing Natural Language Processing (NLP) and machine learning techniques. The system integrates both traditional and modern approaches: a Naive Bayes classifier using TF-IDF for lightweight tasks and a BERT-based model for deep semantic understanding, achieving up to 90% classification accuracy. A robust preprocessing pipeline is implemented, including text cleaning, tokenization, lemmatization, and OCR support for scanned documents. The system accepts multiple formats such as .txt, .pdf, .docx, and image files (.jpg, .png). A user-friendly GUI built with Tkinter allows users to upload files, choose preprocessing steps, select models, and view classification results. The GUI also features visual insights like category-wise word clouds, confusion matrices, sentiment analysis, and keyword charts. The system architecture, UML diagrams, and data flow diagrams are documented to ensure a clear understanding of the workflow. The proposed solution demonstrates that combining classical machine learning with deep learning models significantly enhances legal document classification accuracy and usability, making it a valuable tool for legal firms, courts, and researchers. Keywords:NLP, BERT, Legal Document Classification, Deep Learning, TF-IDF, OCR, GUI, Python, LawTech