Exploiting Contrastive Learning and Numerical Evidence for Confusing Legal Judgment Prediction

Abstract

Given the fact description text of a legal case, legal judgment prediction (LJP) aims to predict the case's charge, applicable law article, and term of penalty. A core challenge of LJP is distinguishing between confusing legal cases that exhibit only subtle textual or number differences. To tackle this challenge, in this paper, we present a framework that leverages MoCo-based supervised contrastive learning and weakly supervised numerical evidence for confusing LJP. Firstly, to make the extraction of numerical evidence (the total crime amount) easier, the framework proposes to formalize it as a named entity recognition task. Secondly, the framework introduces the MoCo-based supervised contrastive learning for multi-task LJP and explores the best strategy to construct positive example pairs to benefit all three subtasks of LJP simultaneously. Extensive experiments on real-world datasets show that the proposed method achieves new state-of-the-art results, particularly for confusing legal cases. Additionally, ablation studies demonstrate the effectiveness of each component.