Retrieval-Augmented Generation and Knowledge Graphs in Portuguese-Language Legal Documents

2026-04-12 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A Graph Retrieval-Augmented Generation (GraphRAG) pipeline has been developed for Question Answering (QA) in Portuguese legal documents. This system was applied to a corpus of 203 normative resolutions from Companhia Energética de Minas Gerais (CEMIG), addressing the inherent structural complexity of legal texts, including hierarchical dependencies and temporal modifications. The approach models documents as knowledge graphs, where nodes represent structural units like Articles, Paragraphs, and Items, and edges denote normative relationships, thereby preserving context and traceability. Its retrieval mechanism reconstructs evidence paths from root to leaf, followed by semantic re-ranking before answer generation. Evaluated using the RAGAS framework, the system achieved a mean answer accuracy of 0.81 and a median of 1.00. The system demonstrates robust performance on short, focused queries, though intermediate-length questions pose challenges due to semantic dispersion.

Key takeaway

For research scientists developing legal AI systems, this GraphRAG pipeline offers a robust method for handling complex Portuguese legal documents. You should consider implementing knowledge graph representations to explicitly model structural dependencies and temporal modifications, which significantly improves interpretability and precision in QA. Be aware that while short queries perform well, you may need to refine semantic re-ranking for intermediate-length questions to mitigate semantic dispersion.

Key insights

GraphRAG enhances legal QA by modeling document structure as knowledge graphs for context-aware retrieval.

Principles

• Explicitly model legal text structure.
• Preserve context via knowledge graph edges.
• Reconstruct evidence paths for traceability.

Method

Documents are modeled as knowledge graphs with structural units as nodes and normative relationships as edges. Retrieval reconstructs evidence paths, followed by semantic re-ranking before generation.

In practice

• Apply GraphRAG to complex, structured documents.
• Use RAGAS for QA system evaluation.
• Focus on short queries for higher accuracy.

Topics

• Graph Retrieval-Augmented Generation
• Portuguese Legal Documents
• Knowledge Graphs
• Question Answering
• RAGAS Framework

Best for: Research Scientist, AI Scientist, NLP Engineer, Legal Professional

Related on AIssential

• Legal-DC: Benchmarking Retrieval-Augmented Generation for Legal Documents — Legal-DC and LegRAG advance Chinese legal RAG by providing specialized benchmarks and a framework for structured legal document processing.
• Combining Semantic Embeddings and Knowledge Graphs for Identifying Decision Patterns in Brazilian Judicial Decisions — Hybrid representations combining semantic embeddings and knowledge graphs enhance judicial decision pattern identification.
• Forget GraphRAG: A 4B AI does the work NOW — Agent-native knowledge orchestration via K1's 4B LLM and multimodal graph overcomes GraphRAG limitations for scientific reasoning.
• Retrieval-Based Multi-Label Legal Annotation: Extensible, Data-Efficient and Hallucination-Free — Retrieval-based legal annotation offers a data-efficient, hallucination-free alternative to generative LLMs for high-cardinality label spaces.
• Development and Evaluation of a Hybrid Information Retrieval System Applied to the Brazilian Legal Domain — Hybrid IR systems combining traditional and Transformer models can overcome long-text limitations in specialized domains.
• Proxy-Pointer RAG: Solving Entity and Relationship Sprawl in Large Knowledge Graphs — Proxy-Pointer RAG uses structural document context for accurate entity and relationship reconciliation in knowledge graphs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.