We report on an experiment in legal judgement prediction on European Court of Human Rights cases where our model first learns to predict the convention articles allegedly violated by the state from case facts descriptions, and subsequently utilizes that information to predict a finding of a violation by the court. We assess the dependency between these two tasks at the feature and outcome level. Furthermore, we leverage a hierarchical contrastive loss to pull together article specific representations of cases at the higher level level, leading to distinctive article clusters, and further pulls the cases in each article cluster based on their outcome leading to sub-clusters of cases with similar outcomes. Our experiment results demonstrate that, given a static pre-trained encoder, our models produce a small but consistent improvement in prediction performance over single-task and joint models without contrastive loss.

翻译：我们报告了一项针对欧洲人权法院案件的法律判决预测实验。在该实验中，我们的模型首先根据案件事实描述学习预测国家涉嫌违反的公约条款，随后利用该信息预测法院是否认定存在违反行为。我们在特征层面和结果层面评估了这两个任务之间的依赖关系。此外，我们利用分层对比损失函数，在高层级上将具有相同条款表征的案件聚集形成独特的条款聚类，并进一步根据案件结果将每个条款聚类内的案件划分为具有相似结果的子聚类。实验结果表明，在静态预训练编码器的基础上，我们的模型相比于无对比损失的单任务模型和联合模型，在预测性能上取得了微小但一致的提升。