Abstract
This article proposes an adaptation of Graph Retrieval Augmented Generation(Graph RAG) specifically designed for the analysis and comprehension of legalnorms, which are characterized by their predefined hierarchical structure,extensive network of internal and external references and multiple temporalversions. By combining structured knowledge graphs with contextually enrichedtext segments, Graph RAG offers a promising solution to address the inherentcomplexity and vast volume of legal data. The integration of hierarchicalstructure and temporal evolution into knowledge graphs - along with the conceptof comprehensive Text Units - facilitates the construction of richer,interconnected representations of legal knowledge. Through a detailed analysisof Graph RAG and its application to legal norm datasets, this article aims toadvance the field of Artificial Intelligence applied to Law, creatingopportunities for more effective systems in legal research, legislativeanalysis, and decision support.