In this study, a machine learning ensemble strategy was proposed and evaluated to enhance the accuracy of classifying legal issues in three areas of Brazilian law: Labor Law, Family Law, and Consumer Law. The approach combined two natural language processing expert systems, trained on both 'popular' and 'non-popular' language texts, along with a classifier to identify them. The results demonstrated that the ensemble classifier achieved an overall accuracy of 96%, significantly outperforming the individual expert systems. However, the strategy increased computational costs, a circumstance that should be consider when one chose to deploy a system like this.