Elamur Research⌕
☰
home / papers / 2026 / MaTIAS – Machine Translation to Inform Asylum Seekers: final results

MaTIAS – Machine Translation to Inform Asylum Seekers: final results

Summary

Asylum reception centres in Belgium house residents who speak many different languages, so staff often struggle to share urgent practical information. MaTIAS, built at Ghent University with Fedasil, addresses this with a one-way notification tool. Staff write short messages in Dutch, French or English, and the messages are machine-translated by ModernMT (adapted with a domain translation memory) into 14 target languages and delivered to residents via WhatsApp.

This paper covers the project's final phase. The prototype ran on Fedasil servers in seven centres, rolled out gradually with demos, training and a support channel for each centre. Staff interviews and surveys showed the tool was considered easy to use, mainly for organisational notices, and residents responded well, though not everyone read the messages. Two expert evaluations found large differences in translation quality across languages. Tigrinya output was judged unusable.

The work shows both the practical value of MT-based messaging for vulnerable groups and its limits for low-resource languages. It also highlights requirements for scaling up: workflow integration, GDPR compliance and usage protocols.

Key points

  • The platform translates from Dutch, French or English into 14 languages chosen by Fedasil. It syncs resident data with Fedasil's database every 24 hours. The project ran 2023–2025 and was co-funded by AMIF (093-133).
  • Deployment covered seven centres across Flanders, Brussels and Wallonia, with demos held in October–November 2025. Feedback came from 13 interviewees and 9 Qualtrics respondents.
  • The most frequent target languages in practice were French, English, Tigrinya, Turkish, Arabic and Somali.
  • MT evaluation 1 (May–June 2025) used 90 messages, all 14 languages and 28 experts. Tigrinya, Somali and Persian were weak.
  • MT evaluation 2 (March 2026) used 40 real messages per language for Tigrinya, Arabic, Somali, Pashto and Persian, each rated by one expert:
    • Arabic, Persian and Somali were acceptable but sometimes hard to follow.
    • Pashto was too formal.
    • Tigrinya often lost the core meaning.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages · Machine translation

Topics: Machine translation

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{wilde2026matias,
  title = {{MaTIAS – Machine Translation to Inform Asylum Seekers: final results}},
  author = {Wilde, July De and Wouters, Anaïs and Tezcan, Arda and Meersschaut, Simon Van den and Maryns, Katrijn and Macken, Lieve},
  booktitle = {Proceedings of the 26th Annual Conference of the European Association for Machine Translation (Volume 2)},
  year = {2026},
  url = {https://aclanthology.org/2026.eamt-2.2/}
}

References

3 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Macken, Lieve, Margot Fonteyne, Arda Tezcan, Ella van Hest, Katrijn Maryns, and July De Wilde. 2025. Machine translation in asylum reception centres : system selection and multilingual quality evaluation. Revista Tradumàtica. Tecnologies de la Traducció, (23):326–349. find
  2. Moerman, Thomas, Arda Tezcan, and Lieve Macken. 2026. Multilingual Communication in the Asylum Context: Evaluating LLM-Based Machine Translation with Fuzzy-Match Augmentation and Adaptive NMT across Resource Conditions under Low-Data Constraints. In Proceedings of the 26th Annual Conference of the E… find our card
  3. See ModernMT’s privacy policy: modernmt.com/privacy https://www. find

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wilde-2026-matias-machine-translation-inform-asylum/