Towards Developing an Automated Chatbot for Predicting Legal Case Outcomes: A Deep Learning Approach

The accurate prediction of legal case outcomes is crucial for effective legal advocacy, which relies on a deep understanding of past cases. Our research aims to develop an automated chatbot for predicting the outcomes of employment-related legal cases using deep learning techniques. We compare and significantly improve on mining the New Zealand Employment Relations Authority (NZERA) dataset, using various deep learning models such as Latent Dirichlet Allocation (LDA) with different activation functions of Recurrent Neural Network (RNN) to determine their predictive performance. Our study's findings show that SoftSign-based RNN-LDA models have the highest accuracy and consistency in predicting outcomes.