AutoLAW: Augmented Legal Reasoning through Legal Precedent Prediction
This addresses an unmet need in the legal community by potentially increasing access to justice through automated legal reasoning.
The paper tackles the problem of predicting relevant legal precedents for legal arguments by introducing Legal Precedent Prediction (LPP) and demonstrates a BERT model that achieves 96% top-10 accuracy on test examples and successfully predicts precedents for a brief by a Supreme Court Justice.
This paper demonstrate how NLP can be used to address an unmet need of the legal community and increase access to justice. The paper introduces Legal Precedent Prediction (LPP), the task of predicting relevant passages from precedential court decisions given the context of a legal argument. To this end, the paper showcases a BERT model, trained on 530,000 examples of legal arguments made by U.S. federal judges, to predict relevant passages from precedential court decisions given the context of a legal argument. In 96% of unseen test examples the correct target passage is among the top-10 predicted passages. The same model is able to predict relevant precedent given a short summary of a complex and unseen legal brief, predicting the precedent that was actually cited by the brief's co-author, former U.S. Solicitor General and current U.S. Supreme Court Justice Elena Kagan.