RAG construction method for heterogeneous graph-hierarchical text corpus analysis

Ulizko M.V., Beresnev A.D., Zhukov V.V., Markov P.D., Gusarova N.F. RAG construction method for heterogeneous graph-hierarchical text corpus analysis. Scientific and Technical Journal of Information Technologies, Mechanics and Optics, 2026, vol. 26, no. 4, pp. 860–867 (in Russian). doi: 10.17586/2226-1494-2026-26-4-860-867

Analysis of document corpora with complex internal and inter-document links requires not only retrieval of relevant texts but also construction of a compact, verifiable and traceable set of fragments sufficient for downstream reasoning and citation. The purpose of this work is to propose a fragment retrieval method for document corpora in which the meaning of a fragment is determined by its local content, position in the document hierarchy and links to other fragments. The method is based on a joint representation of the corpus as a tree of structural units and a directed link graph as well as on hybrid ranking that combines lexical search, vector similarity, and a link signal. The monotonicity and submodularity of the objective function are shown, which makes it possible to use greedy algorithms with a known approximation guarantee and to perform budgeted context selection for a Retrieval-Augmented Generation (RAG) system. In addition, an evaluation protocol is introduced that separates retrieval quality at the document, fragment, and citation levels. The method is formally validated on tests of lexical mismatch robustness and budgeted selection efficiency compared with simple strategies. Examples from the legal domain are used to illustrate the method. The method can be used as a retrieval layer for RAG systems in question answering, evidence retrieval, regulatory compliance, and analysis of large structurally connected corpora.

1. Gao Y., Xiong Y., Gao X., Jia K., Pan J., Bi Y., et al. Retrieval-augmented generation for Large Language Models: a survey. arXiv, 2024. arXiv:2312.10997. doi: 10.48550/arXiv.2312.10997

2. Barnett S., Kurniawan S., Thudumu S., Brannelly Z., Abdelrazek M. Seven failure points when engineering a retrieval augmented generation system. Proc. of the IEEE/ACM 3^rd International Conference on AI Engineering - Software Engineering for AI, 2024, pp. 194–199. doi: 10.1145/3644815.3644945

3. Jin B., Yoon J., Han J., Arik S.O. Long-context LLMs meet RAG: overcoming challenges for long inputs in RAG. Proc. of the International Conference on Learning Representations, 2025.

4. Lewis P., Perez E., Piktus A., Petroni F., Karpukhin V., Goyal N., et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. Proc. of the 34^th International Conference on Neural Information Processing Systems, 2020, pp. 9459–9474.

5. Pipitone N., Alami G.H. LegalBench-RAG: a benchmark for retrieval-augmented generation in the legal domain. arXiv, 2024. arXiv:2408.10343. doi: 10.48550/arXiv.2408.10343

6. Reuter M., Lingenberg T., Liepina R., Lagioia F., Lippi M., Sartor G., et al. Towards reliable retrieval in RAG systems for large legal datasets. Proc. of the Natural Legal Language Processing Workshop, 2025, pp. 17–30. doi: 10.18653/v1/2025.nllp-1.3

7. Zheng L., Guha N., Arifov J., Zhang S., Skreta M., Manning C.D., et al. A Reasoning-focused legal retrieval benchmark. Proc. of the 2025 Symposium on Computer Science and Law, 2025, pp. 169–193. doi: 10.1145/3709025.3712219

8. Latif S., Ameer H., Akram M.H., Fatima M. The chunking paradigm: pecursive semantic for RAG optimization. Proc. of the 8^th International Conference on Natural Language and Speech Processing (ICNLSP-2025), 2025, pp. 137–145.

9. Gunther M., Mohr I., Williams D.J., Wang B., Xiao H. Late chunking: contextual chunk embeddings using long-context embedding models. arXiv, 2024. arXiv:2409.04701. doi: 10.48550/arXiv.2409.04701

10. Qu R., Tu R., Bao F.S. Is semantic chunking worth the computational cost? Findings of the Association for Computational Linguistics: NAACL, 2025, pp. 2155–2177. doi: 10.18653/v1/2025.findings-naacl.114

11. Shin J., Park C., Park J., Seo J., Lim H. MultiDocFusion: hierarchical and multimodal chunking pipeline for enhanced RAG on long industrial documents. Proc. of the Conference on Empirical Methods in Natural Language Processing, 2025, pp. 20985–21004. doi: 10.18653/v1/2025.emnlp-main.1062

12. Sarthi P., Abdullah S., Tuli A., Khanna S., Goldie A., Manning C.D. RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval // Proc. of the International Conference on Learning Representations, 2024.

13. Robertson S., Zaragoza H. The Probabilistic relevance framework: BM25 and beyond. Foundations and Trends in Information Retrieval, 2009. vol. 3, no. 4, pp. 333–389. doi: 10.1561/1500000019

14. Santhanam K., Khattab O., Saad-Falcon J., Potts C., Zaharia M. ColBERTv2: Effective and efficient retrieval via lightweight late interaction. Proc. of the Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2022, pp. 3715–3734. doi: 10.18653/v1/2022.naacl-main.272

15. Khuller S., Moss A., Naor J. The budgeted maximum coverage problem. Information Processing Letters, 1999, vol. 70, no. 1, pp. 39–45. doi: 10.1016/S0020-0190(99)00031-9

16. Lin H., Bilmes J. A class of submodular functions for document summarization. Proc. of the 49^th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, 2011, pp. 510–520.

17. Sviridenko M. A note on maximizing a submodular set function subject to a knapsack constraint. Operations Research Letters, 2004, vol. 32, no. 1, pp. 41–43. doi: 10.1016/S0167-6377(03)00062-2

18. Saveliev D., Kuchakov R. The Russian legislative corpus. arXiv, 2024. arXiv:2406.04855. doi: 10.48550/arXiv.2406.04855

19. Koreeda Y., Manning C. ContractNLI: a dataset for document-level natural language inference for contracts. Findings of the Association for Computational Linguistics: EMNLP, 2021, pp. 1907–1919. doi: 10.18653/v1/2021.findings-emnlp.164

20. Wang S.H., Scardigli A., Tang L., Chen W., Levkin D., Chen A., et al. MAUD: an expert-annotated legal NLP dataset for merger agreement understanding. Proc. of the Conference on Empirical Methods in Natural Language Processing, 2023, pp. 16369–16382. doi: 10.18653/v1/2023.emnlp-main.1019

21. Hendrycks D., Burns C., Chen A., Ball S. CUAD: an expert-annotated NLP dataset for legal contract review. arXiv, 2021. arXiv:2103.06268.doi: 10.48550/arXiv.2103.06268

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License