Evaluating Machine Translation and Automatic Metrics in Subtitling: A Case Study on Spanish Multiword Expressions

María Miró Maestre, Iván Martínez-Murillo


Abstract
Evaluating the translation of multi-word expressions (MWEs) remains a major challenge for Machine Translation (MT), particularly in audiovisual subtitling, where idiomatic meaning and cultural context are essential for adequacy. This study investigates both the ability of state-of-the-art MT systems to translate Spanish MWEs into English and the extent to which current automatic evaluation methods reflect expert human judgment. We introduce ALMO-MWE, a dataset of 235 MWEs extracted from four films by Pedro Almodóvar to evaluate four MT systems using automatic metrics, LLM-as-a-judge approaches, and professional human assessment. Our results reveal a substantial mismatch between traditional automatic metrics and human judgments: n-gram-based metrics show near-zero correlation with expert evaluation and only limited discriminative capacity. In contrast, neural metrics and LLM-based judges exhibit substantially stronger alignment with human assessments, with GPT-OSS achieving the highest overall correlation. These findings highlight fundamental limitations of surface-form metrics for culturally and contextually sensitive translation phenomena and underscore the need for context-aware evaluation frameworks when assessing the translation quality of MWEs in audiovisual translation.
Anthology ID:
2026.eamt-1.29
Volume:
Proceedings of the 26th Annual Conference of the European Association for Machine Translation (Volume 1)
Month:
June
Year:
2026
Address:
Tilburg, The Netherlands
Editors:
Dimitar Shterionov, Eva Vanmassenhove, Mirella De Sisto, Fred Blain, Javad Pourmostafa Roshan Sharami, Lisa Lepp, Chiara Manna, Argentina Anna Rescigno, Alina Karakanta, Ayla Rigouts Terryn, Manuel Lardelli, Natalia Resende, Elena Murgolo, Janiça Hackenbuchner, Anna Zaretskaya, Miquel Esplà-Gomis, Thierry Etchegoyhen, Dagmar Gromann, Rachel Bawden, Barry Haddow, Sara Szoc, Mikel Forcada, Helena Moniz
Venue:
EAMT
SIG:
Publisher:
European Association for Machine Translation
Note:
Pages:
459–472
Language:
URL:
https://aclanthology.org/2026.eamt-1.29/
DOI:
Bibkey:
Cite (ACL):
María Miró Maestre and Iván Martínez-Murillo. 2026. Evaluating Machine Translation and Automatic Metrics in Subtitling: A Case Study on Spanish Multiword Expressions. In Proceedings of the 26th Annual Conference of the European Association for Machine Translation (Volume 1), pages 459–472, Tilburg, The Netherlands. European Association for Machine Translation.
Cite (Informal):
Evaluating Machine Translation and Automatic Metrics in Subtitling: A Case Study on Spanish Multiword Expressions (Miró Maestre & Martínez-Murillo, EAMT 2026)
Copy Citation:
PDF:
https://aclanthology.org/2026.eamt-1.29.pdf