EVALUATING RETRIEVAL-AUGMENTED GENERATION (RAG) SYSTEMS FOR UZBEK LEGAL QUESTION ANSWERING: ARCHITECTURAL FRAMEWORKS, EVALUATION METRICS, AND EMPIRICAL PERFORMANCE
DOI:
https://doi.org/10.37547/Abstract
Retrieval-Augmented Generation (RAG) has emerged as a cornerstone architecture for grounding Large Language Models (LLMs) in domain-specific, authoritative corpora. However, deploying RAG systems within the domain of Uzbek legal question answering presents distinct computational, morphological, and structural challenges. The Uzbek language is a low-resource, morphologically rich, highly agglutinative Turkic language characterized by complex suffixation, dual-script usage (Latin and Cyrillic), and non-trivial semantic parsing requirements. Concurrently, Uzbek statutory frameworks—comprising the Constitution, codes (e.g., Civil, Criminal, Tax Codes), national acts (Qonunlar), and executive decrees (Qarorlar)—exhibit rigid hierarchical structures (document -> section -> chapter -> article -> paragraph) where legal validity is highly sensitive to cross-referential dependencies and temporal amendments.
Downloads
References
1.Esenov, A., Kalandarov, U., & Rakhimov, N. (2024). Natural Language Processing for Turkic Languages: Morphological Parsing and Dense Retrieval Challenges. Central Asian Journal of Computer Science, 8(2), 112–128.
2.Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. Advances in Neural Information Processing Systems (NeurIPS), 33, 9459–9474.
3.Nurali, S. (2025). Uzbek Legal Corpus: A Structured Dataset for Hierarchical Legal Retrieval and QA. Hugging Face Datasets.
https://huggingface.co/datasets/sukhrobnurali/uzbek-legal-corpus
4.Oro, E. (2024). Evaluating Retrieval-Augmented Generation for Question Answering with Large Language Models. CEUR Workshop Proceedings, 3762, 495–508.
5.Shah, C., & Bender, E. M. (2022). Situating Search: Issues with Using Large Language Models as Search Engines. ACM Transactions on Information Systems, 40(4), 1–28.
5.Toloka AI. (2025). RAG evaluation: a technical guide to measuring retrieval-augmented generation. Toloka Engineering Reports. https://toloka.ai/blog/rag-evaluation-a-technical-guide-to-measuring-retrieval-augmented-generation/
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Articles published in this journal are licensed under the Creative Commons Attribution 4.0 International License (CC-BY 4.0). Under this license:
- Share: Copy and redistribute the material in any medium or format
- Adapt: Remix, transform, and build upon the material for any purpose, including commercially
Attribution required: You must give appropriate credit, provide a link to the license, and indicate if changes were made.
License URL: https://creativecommons.org/licenses/by/4.0/
Authors retain copyright of their work while granting the journal first publication rights.
