Using Natural Language Processing for Case Brief Generation and Verdict Support in the Kenyan Court System

Abstract

The National Council for Law Reporting publishes judgements in an online searchable database, enabling large-scale machine learning and statistical analysis in the legal domain. This is the culmination of the transformation that had been going on in the Judiciary inspired by the new requirements of public service delivery under the 2010 Constitution and the increased awareness and demand for legal information by the citizens. The Kenyan Judiciary is now continually seeking to apply creative, innovative, appropriate and integrated technological solutions that enable efficient service delivery. This research focuses on how to integrate Natural Language Processing (NLP) systems and Artificial Intelligence (AI) in document review, legal writing and legal case predictions in Kenya leveraging on NLP’s major purpose which is converting informal textual structures into formal representations for analysis. The aim is to demonstrate that NLP and Machine Learning (ML) algorithms can be exploited to provide a viable means of solving the problems bedeviling the Kenyan judicial system such as the increasing complexity of cases and huge backlog of cases in Kenyans courts. Using selected lawyers and advocates to provide expert labeling of the downloaded cases and sample legal case briefs to fine tune and evaluate the outcome of the summary models, NLP and AI algorithms were used to automatically generate case briefs with relevant precedent cases and the likely outcome of the verdict associated with the case submitted to the Kenyan Judiciary. The toolkit developed is a trained NLP and AI model that can generate case briefs and predetermined verdicts of the specific case with 88% and 83.7% levels of accuracy respectively. Considering the huge backlog of cases in Kenyans courts, coupled with the complexity of the cases, this research has demonstrated that NLP and ML can augment human abilities and provide a viable means of automating some aspects of the legal process such as case brief generation and verdict prediction. The toolkit developed, when fully implemented, will result in improving service delivery through facilitating speedier trials and enhancing the efficiency and effectiveness of administrative processes.