Interpretation of idioms and culturally marked elements in neural machine translation (based on Russian-English parallels)

Cover Page

Cite item

Full Text

Abstract

The article is dedicated to the study of the features of interpreting idioms and culturally marked elements in machine translation from Russian to English using neural network systems such as ChatGPT, DeepL, and Google Translate. The focus is on the problem of preserving the idiomatic meaning and cultural connotations, as well as the degree of cognitive adaptation of artificial intelligence to nationally specific realities. The relevance of this study is determined by the fact that in the era of global communication and digital interaction, language models become tools not only for interlingual exchange but also for cultural mediation. The question of whether artificial intelligence is capable of conveying the cultural meanings behind idioms, realities, and metaphors has fundamental significance for both translation theory and cognitive linguistics. The author conducts a corpus-based comparison of translations of real idioms and fixed expressions, identifying typical errors such as calque, semantic alignment, loss of imagery, and pragmatic meaning. Based on the analysis, a classification of translation strategies used by neural network systems is proposed: literal translation, functional adaptation, and cognitive substitution. The theoretical significance of the work lies in deepening the understanding of the mechanisms of interaction between cognitive and cultural factors in machine translation. The practical significance lies in the possibility of applying the results obtained in the development of multimodal translation models aimed at preserving the national-cultural component. It is concluded that despite the high level of syntactic and lexical accuracy, neural network models do not provide full cultural equivalence and require the integration of cognitive-cultural corpora into the training process. The theoretical implications of the results consist in clarifying the boundary between "formal equivalence" and "cognitive-cultural adequacy" in the context of machine translation. Neural network models, even with access to extensive discursive data, still imitate rather than reproduce cultural memory and empathetic understanding necessary for a sustainable interpretation of figurative units. This means that progress in the field of neuro-translation is directly related to the inclusion of specialized cultural corpora, knowledge graphs, and annotated phraseological resources in training, as well as the development of mechanisms for global contextual attention capable of considering not only adjacent sentences but also genre-rhetorical conventions of the text.

References

  1. Комиссаров В. Н. Теория перевода: лингвистические аспекты. – М.: Наука, 1990.
  2. Швейцер А. Д. Теория перевода: статус, проблемы, аспекты. – М.: Наука, 1988.
  3. Рецкер Ю. И. Теория перевода и переводческая практика. – М.: Международные отношения, 1974.
  4. Найда Ю. (Nida E.) Прагматическая и лингвистическая эквивалентность. – М.: Прогресс, 1980.
  5. Бейкер М. (Baker M.) Перевод и лексика: учебное пособие In Other Words. – Лондон: Routledge, 1992.
  6. Хэтим Б., Мейсон И. (Hatim B., Mason I.) Дискурс и перевод. – Лондон: Longman, 1990.
  7. Лакофф Дж., Джонсон М. (Lakoff G., Johnson M.) Метафоры, которыми мы живём. – Чикаго: University of Chicago Press, 1980.
  8. Бархударов Л. С. Язык и перевод. – М.: Международные отношения, 1975.
  9. Toral A., Way A. What Level of Quality can Neural Machine Translation Attain? // Machine Translation. – 2018. – Т. 32(2). – С. 179-193.
  10. Freitag M., Bapna A., Firat O. et al. Experts, Errors, and Context in Neural Machine Translation. // TACL. – 2022. – Т. 10. – С. 20-37.
  11. Базиотис К., Матор П., Хаслер Э. (Baziotis C., Mathur P., Hasler E.) Automatic Evaluation and Analysis of Idioms in Neural Machine Translation // Proceedings of EACL. – 2023.
  12. Навин П., Троёвский П. (Naveen P., Trojovský P.) Overview and Challenges of Machine Translation for Contextually Appropriate Translations // iScience. – 2024. – Vol. 27, Issue 10. doi: 10.1016/j.isci.2024.110878 EDN: NOZVGM
  13. Лю Э., Чодхари А., Ньюбиг Г. (Liu E., Chaudhary A., Neubig G.) Crossing the Threshold: Idiomatic Machine Translation through Retrieval Augmentation and Loss Weighting // EMNLP. – 2023.
  14. Кастальдо А., Монти Д. (Castaldo A., Monti J.) Prompting Large Language Models for Idiomatic Translation // Creative-text Translation and Technology Workshop. – 2024.
  15. Азизов Д. (Azizov D.) From Idioms to Algorithms: Translating Culture-Specific Expressions in AI Systems // IRE Journals. – 2024. – Vol. 7, No. 10. – С. 543-551.
  16. Дарвиш Н. и др. (Darwish N. et al.) A Reception Study of AI-Translated Idioms and Proverbs Between Arabic and English // Research Journal in Advanced Humanities. – 2025. – Т. 6(3).
  17. Альделаа А. С. (Aldelaa A. S.) Investigating Problems Related to the Translation of Idiomatic Expressions in the Arabic Novels Using Neural Machine Translation // Theory and Practice in Language Studies. – 2024. – Т. 14(1). – С. 71-78. doi: 10.17507/tpls.1401.08 EDN: TXCMBF.

Supplementary files

Supplementary Files
Action
1. JATS XML
Download

Согласие на обработку персональных данных

 

Используя сайт https://journals.rcsi.science, я (далее – «Пользователь» или «Субъект персональных данных») даю согласие на обработку персональных данных на этом сайте (текст Согласия) и на обработку персональных данных с помощью сервиса «Яндекс.Метрика» (текст Согласия).