Haiku. A Comparative Analysis of Human and AI Translation

The rapid development of neural machine translation systems raises the question of their applicability to literary texts, especially poetry, which is characterised by a high degree of semantic compression and cultural conditioning. One of the most difficult poetic genres to translate is haiku, in which meaning is formed through conciseness, imagery, cultural markers, and the principles of understatement. The number of studies in the field of automated translation is growing, but comparative analysis of haiku translations performed by humans and artificial intelligence systems remains underdeveloped. Objectives. The aim of the research is to identify the characteristics and limitations of translations of Japanese haiku by artificial intelligence systems in comparison with translations by humans. Methods. The study was conducted in the form of a qualitative comparative analysis. The material consisted of seven haiku by Matsuo Basho, for which five translation options were considered: two performed by humans and three using modern neural machine translation systems. The analysis was conducted according to a number of criteria: preservation of the figurative structure, transmission of cultural realities, degree of explication of the original meaning, reproduction of poetic conciseness, and the presence of interpretative additions. Results. It was found that translations performed by artificial intelligence tend to emphasize semantic explanation and reduce the level of figurative ambiguity. In a number of cases, there is a loss of culturally specific elements and a shift in emphasis, which leads to a change in the aesthetic effect of the text. Translations performed by humans demonstrate greater variability in interpretation while preserving the genre and cultural characteristics of haiku. Conclusions. The results confirm that modern artificial intelligence systems are effective in conveying general content but have difficulty reproducing poetic ambiguity and cultural implicatures. Prospects for further research are linked to the expansion of the text corpus and the development of hybrid translation models that combine automated and expert approaches.

1. Pignocchi T. Ordinary aesthetics and ethics in the Haiku poetry of Matsuo Bashō: A Wittgensteinian perspective. Open Philosophy. 2023;6(1). URL: https://doi.org/10.1515/opphil‑2022-0255

2. Meilantari N., Suarka I., Sudewa I., Triadnyani I. The evolution of Japanese Haiku: From paper to digital. International Journal of Education and Literature. 2024;3:128-135. URL: https://doi.org/10.55606/ijel.v3i2.118

3. Dyakonova E.M. Poetry of the Japanese three-verse genre (haiku): Origin and main features. In collection: Lyrics: Genesis and evolution. Moscow: Publishing House of RGGU; 2007:337-352. (In Russ.).

4. Gerasimova I.A., Ivanov A.V., Fotieva I.V. Transcendental experience: The possibility of intersection of theological, philosophical and scientific approaches. Culture and Art. 2023;3:9-24. (In Russ.). URL: https://doi.org/10.7256/2454-0625.2023.3.38068

5. Zlygostev V.Y. The world of hokku. “Life between the Lines”: The philosophy of Understatement in Japanese Haiku poetry. A symbol of science. 2022;8-1:55-61. URL: https://os-russia.com/SBORNIKI/SN‑2022-08-1. pdf?ysclid=mopxcmsqef240436089 (In Russ.).

6. Dolin A. Poetry and poetics of Haiku. Bulletin of Cultural Studies. 2011;3(58):111-121. URL: https://www.elibrary.ru/obgclp (In Russ.).

7. Norets M.V., Reinova A.V. Artistic translation and artificial intelligence: Tradition and innovation. International Scientific Research Journal. 2024;12(150):155. (In Russ.). URL: https://doi.org/10.60797/IRJ.2024.150.111

8. Drobot I.-A. What could we understand by translation? Haiku poems’ translations in the Romanian language. In: Studies in Translation and Language. 2024:25-26. URL: https://doi.org/10.5772/intechopen.1002949

9. Hirata K., Yokoyama S., Yamashita T., Kawamura H. Implementation of autoregressive language models for generation of seasonal fixed-form Haiku in Japanese. Knowledge, Information and Creativity Support Systems. 2023;2(1):19-25. URL: https://doi.org/10.52731/liir.v003.075

10. Ozyumenko V.I., Larina T.V. Artificial intelligence in translation: Strengths and weaknesses. Bulletin of the Volgograd State University. Series 2: Linguistics. 2025;24(1):117-130. (In Russ.). URL: https://doi.org/10.15688/jvolsu2.2025.1.10

11. Kolin K.K., Khoroshilov A.A., Nikitin Yu.V., Pshenichny S.I., Khoroshilov A.A. Artificial intelligence in machine translation technologies. Social innovations and social sciences. 2021;2(4):64-80. (In Russ.). URL: https://doi.org/10.31249/snsn/2021.02.05

12. Rarenko M.B. Modern challenges of machine translation. Social sciences and humanities. Russian and foreign literature. Series 6. Linguistics: Abstract Journal. 2022;1:46-53. (In Russ.). URL: https://doi.org/10.31249/ling/2022.01.04

13. Matsuo Basho. “Haiku 1-100”. Transl. from Japanese by D.N. Smirnov-Sadovsky, comm. by I. Yasuda. St. Petreburg: Krasny Matros; 2011. 128 p. (In Russ.).

14. Shelestyuk E.V., Gritsenko E.D. On forenization and domestication in translation and the possibilities of their linguistic assessment. Bulletin of the Chelyabinsk State University. 2016;4(386):202-207. URL: https://www.elibrary. ru/wbkisp (In Russ.).

15. Alekseitseva T.A. Explication or the infinite generation of meaning. Bulletin of the Moscow University. Episode 22: Theory of Translation. 2009;(4):5-15. URL: https://www.elibrary.ru/kyyyox (In Russ.).

16. Shenderyuk-Zhidkov A.V., Maksimenko V.A., Khramov A.E. Coevolution of neurotechnologies and AI: Ethical challenges and approaches to regulation. News of higher educational institutions. Applied nonlinear dynamics. 2026;34(1):116-160. (In Russ.). URL: https://doi.org/10.18500/0869-6632-003196