• Bert
• graph attention networks
• graph convolutional networks
• graph neural networks
• knowledge distillation
• legal classifier
• transformers

Automated text classification is a task of high relevance in digital applications to the legal field. In this thesis, we addressed the problem of developing an automated legal text classifier leveraging a blend of machine learning techniques including graph neural networks and knowledge distillation. The resulting system can help streamline the legal document review process, reducing the time and effort required for manual categorization and improving the accuracy and consistency of the results. The challenges and limitations of building an automated legal text classifier, including the need for high-quality annotated data and the complexities of the legal domain, are also discussed.