Keywords
Abstract
This thesis presents a detailed analysis of cultural and stylistic errors in translations produced by Large Language Models (LLMs). While modern artificial intelligence systems demonstrate high fluency and grammatical correctness, they systematically fail to handle complex linguistic layers such as localized idioms, social registers, and historical contexts. Using advanced translation frameworks like Multidimensional Quality Metrics (MQM), this research explores the limits of AI systems when dealing with cultural nuances. The findings reveal that automated evaluation tools often miss these deep errors, making expert human post-editing indispensable. Ultimately, the study suggests that combining technical solutions like Retrieval-Augmented Generation (RAG) with human expertise is essential to protect cultural identity and maintain stylistic accuracy in digital translation.
References
1. Artetxe, M., Ruder, S., & Conneau, A. (2020). Cross-lingual retrieval for iterative self-supervised training. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2311–2317.
2. Ataman, D., Birch, A., Habash, N., Federico, M., Koehn, P., & Cho, K. (2025). Machine translation in the era of large language models: A survey of historical and emerging problems. Information, 16(9), 723.
3. Plaza, J., Martinez, L., & Gomez, F. (2024). Measuring the impact of mistranslation artifacts on multilingual benchmark reliability. International Journal of Linguistics, Literature and Translation, 7(2), 114–125.
4. Smith, A., & Jones, B. (2025). Machine translation vs. human translation: Quality, challenges, and perceptions. ESP Journals (International Journal of English Language and Literature), 3(2), 104–115.
5. Thellmann, K., Stadler, B., Färber, M., & Lehmann, J. (2026). Quantifying the impact of translation errors on multilingual LLM evaluation. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).
6. Wang, H., & Liu, Z. (2025). Enhancing low-resource minority language translation with LLMs and retrieval-augmented generation for cultural nuances. ResearchGate Preprints.
7. Zhang, Y., Beard, R., Hawkins, J., & Chandra, R. (2026). Automated evaluation of LLMs for effective machine translation of Mandarin Chinese to English. arXiv preprint, arXiv:2603.09998.
8. Zhou, X., & Harris, M. (2026). Prompt-induced cultural mediation and its limits: LLM translation of cultural texts. Cogent Arts & Humanities, 13(1), 2631304.