Using attention methods to predict judicial outcomes

TL;DR

This paper explores the use of attention-based neural networks to predict judicial outcomes in Brazilian legal cases, demonstrating the effectiveness of specific models and analyzing key textual features influencing decisions.

Contribution

It introduces a novel application of attention methods to legal outcome prediction and provides insights into important textual features affecting judicial decisions.

Findings

Hierarchical Attention Networks achieved high prediction metrics.

Attention weights identified key words influencing judgments.

Models outperformed traditional classifiers on legal text data.

Abstract

Legal Judgment Prediction is one of the most acclaimed fields for the combined area of NLP, AI, and Law. By legal prediction we mean an intelligent systems capable to predict specific judicial characteristics, such as judicial outcome, a judicial class, predict an specific case. In this research, we have used AI classifiers to predict judicial outcomes in the Brazilian legal system. For this purpose, we developed a text crawler to extract data from the official Brazilian electronic legal systems. These texts formed a dataset of second-degree murder and active corruption cases. We applied different classifiers, such as Support Vector Machines and Neural Networks, to predict judicial outcomes by analyzing textual features from the dataset. Our research showed that Regression Trees, Gated Recurring Units and Hierarchical Attention Networks presented higher metrics for different subsets. As a final goal, we explored the weights of one of the algorithms, the Hierarchical Attention Networks, to find a sample of the most important words used to absolve or convict defendants.