1. КК
  2. All Volumes and Issues
  3. № 2 (48) / 2025
  4. Design of a Machine Translation Service for Variable Data

Design of a Machine Translation Service for Variable Data

КК.
2025;
Volume 48, Number 2
: pp. 155 - 162
https://doi.org/10.32403/2411-3611-2025-2-48-155-162
Authors:
  1. І. V. Malets
1
Lviv Polytechnic National University, Institute of Printing Art and Media Technologies

The paper addresses the problem of designing a machine translation service for variable data that operates at the level of document segments rather than monolithic text. In contemporary publishing, marketing and educational workflows, documents are increasingly represented as sets of structured segments associated with content keys, language attributes and, in many cases, personalization markers inherited from CRM or content management systems. Conventional web­based machine translation services treat such fragments as plain text, which may lead to distortion or loss of personalization markers and break the logical link between translated text and its source fields. As a result, translation resources become fragmented across platforms, translation memory cannot be reused consistently, and additional human effort is required for manual verification and correction. The proposed approach models each publication unit uᵢ as a set of segments Sᵢ = {sᵢj} that are processed by a centralized machine translation service acting as a language layer between web content management systems, CRM platforms and other client applications. For each segment sᵢj the service combines translation memory and a pool of machine translation engines, selecting the actual output tᵢj either from translation memory or from a chosen engine while preserving personalization markers as atomic tokens. The integration structure is implemented via RESTful endpoints for single­segment processing and for batch translation of segment sets. Both endpoints operate on JSON messages that encapsulate content keys, source and target language codes, translation results, information about the source (translation memory or engine) and processing status. A prototype of the service has been developed using the Flask web framework. Experimental use on structured abstracts of a master’s thesis and fragments of personalized messages demonstrates that the service preserves document structure and variable markers, supports consistent mapping of translations to content keys and enables the creation of a unified translation corpus reusable across multiple distribution channels. The results indicate that segment­level integration of a machine translation service provides a feasible basis for centralized management of language resources and automated preparation of multilingual personalized content.

machine translation
variable data
segment­based document model
REST API
personalized content
service architecture
  1. Khoroshun, O. O. (2021). Mashynnyi pereklad: istorychnyi ohliad. Nova filolohiia, 82, 333– 337. DOI: 10.26661/2414-1135-2021-82-53. (in Ukrainian).
  2. Liubymova, S. A. (2024). Problemy avtomatyzovanoho perekladu: konspekt lektsii dlia zdobuvachiv vyshchoi osvity za druhym (mahisterskym) rivnem. Odesa: Derzhavnyi zaklad «Universytet Ushynskoho». (in Ukrainian).
  3. Moisieieva, N., Dzykovych, O., & Shtanko, A. (2023). Mashynnyi pereklad: porivniannia rezultativ ta analiz pomylok DeepL ta Google Translate. Advanced Linguistics, 11, 78–82. DOI: 10.20535/2617-5339.2023.11.277593. (in Ukrainian).
  4. Podvoiska, O., Kozoriz, I., & Honcharenko, N. (2024). Mashynnyi pereklad za dopomohoiu internet resursiv. Molodyi vchenyi, 4(128), 68–71. DOI: 10.32839/2304-5809/2024-4-128-10. (in Ukrainian).
  5. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. In Advances in Neural Information Processing Systems (pp. 5998–6008). DOI: 10.48550/arXiv.1706.03762DOI: 10.48550/arXiv.1409.3215 (in English).
  6. Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to sequence learning with neural networks. In Advances in Neural Information Processing Systems (pp. 3104–3112). DOI: 10.48550/arXiv.1409.3215. (in English).
  7. Tiedemann, J., & Thottingal, S. (2020). OPUS-MT: Building open translation services for low-resource languages. In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation (pp. 479–480). Lisbon. (in English).
  8. Koehn, P. (2020). Neural machine translation. Cambridge: Cambridge University Press. (in English).
  9. Malets I. (2025). Review of Machine Translation Quality Assessment Methods for Personalized Content Preparation. Youth Science: Innovation and Global Challenges, 2. (in English).
  10. Printing Equipment UkrPol. Retrieved from ukrpol.ua/en/equipments.