Advancing Legal Information Extraction: A GNN-Based Retrieval Attractor Framework for Legal PDF Documents
Resumo
A Extração de Informação (EI) em documentos jurídicos em formato PDF permanece desafiadora devido à complexidade estrutural desses textos. Embora Grandes Modelos de Linguagem (LLMs), Redes Neurais em Grafos (GNNs) e Retrieval-Augmented Generation (RAG) tenham avançado o processamento de documentos jurídicos, abordagens de recuperação ainda apresentam sensibilidade a ruídos e desalinhamento contextual. Esta pesquisa de doutorado propõe um framework de recuperação baseado em sistemas dinâmicos guiados por consulta sobre representações em grafos de documentos jurídicos. A abordagem introduz atratores de recuperação, definidos como regiões semânticas estáveis emergentes de propagação iterativa em grafos, e investiga recuperação em representações Euclidianas e Hiperbólicas. Resultados preliminares sugerem que atratores baseados em grafos representam uma direção promissora para extração de informação jurídica.
Palavras-chave:
Legal Information Extraction, Large Language Models (LLMs), Graph Neural Networks (GNNs), Retrieval-Augmented Generation (RAG), Hyperbolic Representations
Referências
Balažević, I., Allen, C., and Hospedales, T. (2019). Multi-relational Poincaré graph embeddings. Advances in Neural Information Processing Systems (NeurIPS).
Bommarito, M. and Katz, D. M. (2022). Machine Learning for Law: Text, Cases, and Analytics. Cambridge University Press.
Chalkidis, I., Fergadiotis, M., Malakasiotis, P., Aletras, N., and Androutsopoulos, I. (2020). Legal-bert: The muppets straight out of law school.
Chami, I., Ying, R., Ré, C., and Leskovec, J. (2019). Hyperbolic graph convolutional neural networks. Advances in Neural Information Processing Systems (NeurIPS).
Chen, W., Han, X., Lin, Y., Zhao, H., Liu, Z., Li, P., Sun, M., and Zhou, J. (2021). Fully hyperbolic neural networks.
Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. (2019). Bert: Pre-training of deep bidirectional transformers for language understanding. pages 4171–4186. Proceedings of NAACL-HLT.
Hale, J. K. (1980). Ordinary Differential Equations. John Wiley and Sons, New York, 2nd edition.
He, N., Madhu, H., Bui, N., Yang, M., and Ying, R. (2025). Hyperbolic deep learning for foundation models: A survey. Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’25).
Jurafsky, D. and Martin, J. H. (2023). Speech and Language Processing. Pearson, 3rd edition.
Li, H., Su, W., Wang, C., Wu, Y., Ai, Q., and Liu, Y. (2023). Thuir@coliee 2023: Incorporating structural knowledge into pre-trained language models for legal case retrieval.
Li, Y., Wang, G., and Yang, Y. (2025). Implicit hypergraph neural networks. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
Manning, C. D., Raghavan, P., and Schütze, H. (2021). Introduction to Information Retrieval. Cambridge University Press, 2nd edition.
Mavromatis, C. and Karypis, G. (2024). Gnn-rag: Graph neural retrieval for large language model reasoning.
Nguyen, H.-T. and Satoh, K. (2024). ConsRAG: Minimize LLM Hallucinations in the Legal Domain.
Singh, S. (2018). Natural language processing for information extraction.
Strogatz, S. H. (2015). Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering. Westview Press, 2nd edition.
Tang, Y., Qiu, R., and Huang, Z. (2025). Uqlegalai@coliee2025: Advancing legal case retrieval with large language models and graph neural networks.
Tang, Y., Qiu, R., Liu, Y., Li, X., and Huang, Z. (2023). Casegnn: Graph neural networks for legal case retrieval with text-attributed graphs.
Tang, Y., Qiu, R., Liu, Y., Li, X., and Huang, Z. (2024a). Casegnn++: Graph contrastive learning for legal case retrieval with graph augmentation.
Tang, Y., Qiu, R., Yin, H., Li, X., and Huang, Z. (2024b). Caselink: Inductive graph learning for legal case retrieval.
Tifrea, A., Becigneul, G., and Ganea, O. (2021). Hyperbolic deep neural networks: A survey. Journal of Machine Learning Research, 22(169):1–42.
Zhang, B., Liang, Y., and Wang, H. (2025). Gfm-rag: Graph foundation model for retrieval augmented generation.
Zhang, W., Liu, X., and Wang, Y. (2022). Advances in relation extraction for natural language processing. Computational Linguistics, 48(2):245–268.
Bommarito, M. and Katz, D. M. (2022). Machine Learning for Law: Text, Cases, and Analytics. Cambridge University Press.
Chalkidis, I., Fergadiotis, M., Malakasiotis, P., Aletras, N., and Androutsopoulos, I. (2020). Legal-bert: The muppets straight out of law school.
Chami, I., Ying, R., Ré, C., and Leskovec, J. (2019). Hyperbolic graph convolutional neural networks. Advances in Neural Information Processing Systems (NeurIPS).
Chen, W., Han, X., Lin, Y., Zhao, H., Liu, Z., Li, P., Sun, M., and Zhou, J. (2021). Fully hyperbolic neural networks.
Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. (2019). Bert: Pre-training of deep bidirectional transformers for language understanding. pages 4171–4186. Proceedings of NAACL-HLT.
Hale, J. K. (1980). Ordinary Differential Equations. John Wiley and Sons, New York, 2nd edition.
He, N., Madhu, H., Bui, N., Yang, M., and Ying, R. (2025). Hyperbolic deep learning for foundation models: A survey. Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’25).
Jurafsky, D. and Martin, J. H. (2023). Speech and Language Processing. Pearson, 3rd edition.
Li, H., Su, W., Wang, C., Wu, Y., Ai, Q., and Liu, Y. (2023). Thuir@coliee 2023: Incorporating structural knowledge into pre-trained language models for legal case retrieval.
Li, Y., Wang, G., and Yang, Y. (2025). Implicit hypergraph neural networks. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).
Manning, C. D., Raghavan, P., and Schütze, H. (2021). Introduction to Information Retrieval. Cambridge University Press, 2nd edition.
Mavromatis, C. and Karypis, G. (2024). Gnn-rag: Graph neural retrieval for large language model reasoning.
Nguyen, H.-T. and Satoh, K. (2024). ConsRAG: Minimize LLM Hallucinations in the Legal Domain.
Singh, S. (2018). Natural language processing for information extraction.
Strogatz, S. H. (2015). Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering. Westview Press, 2nd edition.
Tang, Y., Qiu, R., and Huang, Z. (2025). Uqlegalai@coliee2025: Advancing legal case retrieval with large language models and graph neural networks.
Tang, Y., Qiu, R., Liu, Y., Li, X., and Huang, Z. (2023). Casegnn: Graph neural networks for legal case retrieval with text-attributed graphs.
Tang, Y., Qiu, R., Liu, Y., Li, X., and Huang, Z. (2024a). Casegnn++: Graph contrastive learning for legal case retrieval with graph augmentation.
Tang, Y., Qiu, R., Yin, H., Li, X., and Huang, Z. (2024b). Caselink: Inductive graph learning for legal case retrieval.
Tifrea, A., Becigneul, G., and Ganea, O. (2021). Hyperbolic deep neural networks: A survey. Journal of Machine Learning Research, 22(169):1–42.
Zhang, B., Liang, Y., and Wang, H. (2025). Gfm-rag: Graph foundation model for retrieval augmented generation.
Zhang, W., Liu, X., and Wang, Y. (2022). Advances in relation extraction for natural language processing. Computational Linguistics, 48(2):245–268.
Publicado
08/09/2026
Como Citar
SANTOS, Vitória S.; DORNELES, Carina F..
Advancing Legal Information Extraction: A GNN-Based Retrieval Attractor Framework for Legal PDF Documents. In: WORKSHOP DE TESES E DISSERTAÇÕES (WTDBD) - SIMPÓSIO BRASILEIRO DE BANCO DE DADOS (SBBD), 41. , 2026, São Carlos/SP.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 335-341.
DOI: https://doi.org/10.5753/sbbd_estendido.2026.249679.
