Towards Automated Classification of Repetitive Themes in Brazilian Courts with LegalClass

The growing influx of lawsuits in judicial systems presents a pressing challenge for timely case resolution. The Sao Paulo Justice Court is particularly noteworthy, boasting the world’s largest caseload with an 84% congestion rate and an average processing time of over seven years. To address this issue, this article introduces LegalClass, a computational tool designed to expedite case processing. LegalClass employs natural language processing and an array of machine learning algorithms-such as Support Vector Machines, Logistic Regression, Naive Bayes, K-Nearest Neighbors, and Convolutional Neural Networks-for the automated classification of lawsuits, with a focus on repetitive legal themes. Developed in collaboration with the University of São Paulo, this tool aims to significantly enhance the efficiency of the judicial system by harnessing the capabilities of artificial intelligence and data science. This study uses LegalClass to assess the performance of machine learning algorithms in categorizing lawsuits according to predefined themes set by the Superior Court of Justice. While automation through text classification has shown promise in handling vast volumes of legal texts, ongoing improvements in methods and techniques are essential for increasing both efficiency and accuracy. Continued research in this rapidly evolving field is crucial for meeting the changing needs of legal information processing.