XML and JSON Translator
Khan, Ishup Ali (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025060521005
https://urn.fi/URN:NBN:fi:amk-2025060521005
Tiivistelmä
This thesis explores key challenges in localizing structured XML and JSON documents. These challenges were specifically focused on the need to preserve document structure post localization, reduce reliance on localization vendors, and integrate with existing software ecosystems. A custom machine translation application was developed that utilizes the machine translation services to translate structured documents while maintaining structural integrity.
The translator application was tested across multiple European languages. The application was effective in preserving nested XML structures, HTML tags, and placeholders. Qualitative analysis of these translated content showed that translation quality varied by languages complexities. Two translation services Claude API and Hugging Face Helsinki NLP models were used to translate the structured content. Among the two translation services used, Claude API generally produced more contextually appropriate translations in comparison to the translation from Hug-ging Face Helsinki NLP models. User evaluations highlighted structural preservation and workflow compatibility as the most valuable features of the application.
The implementation leverages a containerized architecture which support flexible deployment across on-premises and cloud environments. This thesis demonstrates that targeted, structure-preserving translation workflows can significantly enhance localization processes by reducing manual intervention and improving consistency. The approach offers a practical alternative to fully human translation and highlights the potential to utilize machine and human translation forming a hybrid workflow in the evolving landscape of language technologies.
The translator application was tested across multiple European languages. The application was effective in preserving nested XML structures, HTML tags, and placeholders. Qualitative analysis of these translated content showed that translation quality varied by languages complexities. Two translation services Claude API and Hugging Face Helsinki NLP models were used to translate the structured content. Among the two translation services used, Claude API generally produced more contextually appropriate translations in comparison to the translation from Hug-ging Face Helsinki NLP models. User evaluations highlighted structural preservation and workflow compatibility as the most valuable features of the application.
The implementation leverages a containerized architecture which support flexible deployment across on-premises and cloud environments. This thesis demonstrates that targeted, structure-preserving translation workflows can significantly enhance localization processes by reducing manual intervention and improving consistency. The approach offers a practical alternative to fully human translation and highlights the potential to utilize machine and human translation forming a hybrid workflow in the evolving landscape of language technologies.