Go to ACL ARR 2024 August homepageNyayaAnumana and INLegalLlama The Largest Indian Legal Judgment Prediction Dataset and Specialized Language Model for Enhanced Decision Analysis
Abstract: The integration of AI in legal judgment prediction (LJP) has the potential to transform the legal landscape, particularly in jurisdictions like India, where the legal system is burdened by a significant backlog of cases. This paper introduces NyayaAnumana, the largest and most diverse corpus of Indian legal cases compiled for LJP, encompassing a total of 702,945 preprocessed cases. NyayaAnumana, which combines the Hindi words "Nyay" (judgment) and "Anuman" (prediction or inference), includes a wide range of cases from the Supreme Court, High Courts, Tribunal Courts, District Courts, and Daily Orders, providing unparalleled diversity and coverage. Our dataset surpasses existing datasets like PredEx and ILDC, offering a comprehensive foundation for advanced AI research in the legal domain. In addition to the dataset, we present INLegalLlama, a domain-specific generative LLM tailored to the intricacies of the Indian legal system. It is developed through a two-phase training approach: injecting legal knowledge and enhancing reasoning capabilities. This method allows the model to achieve a deep understanding of legal contexts. Our experiments demonstrate that incorporating diverse court data significantly boosts model accuracy, achieving approximately 90\% F1 score in prediction tasks. INLegalLlama not only improves prediction accuracy but also offers comprehensible explanations, addressing the need for explainability in AI-assisted legal decisions. These contributions advance both the technological and practical aspects of LJP, highlighting the importance of diverse datasets in developing effective AI solutions for the legal field.
Paper Type: Long
Research Area: Resources and Evaluation
Research Area Keywords: NLP Applications, Resources and Evaluation, Language Modeling, Interpretability and Analysis of Models for NLP
Contribution Types: Model analysis & interpretability, Approaches low compute settings-efficiency, Publicly available software and/or pre-trained models, Data resources
Languages Studied: English
Submission Number: 52
Loading