Classifying parsed legal documents using a voting-based ensemble of Naïve Bayes and support vector machine with hierarchical knowledge graph visualization

Complex legal documents often cause non experts to agree to unfavorable terms due to their intricate terminology. To make these contracts accessible, a web application was created to deconstruct and categorize financial clauses using synthetic data from major Malaysian banks. The system employs an ensemble model combining Naive Bayes and Support Vector Machine algorithms to classify terms across six critical categories, achieving over eighty percent accuracy. Results are displayed through an interactive knowledge graph, concise summaries, and a jargon dictionary, which together reduced analysis time for users by more than seventy percent compared to manual reading. Ultimately, this tool empowers the public to make better financial decisions by transforming complex legal language into a clear, interactive format.