Dataset card
Description
FLORES+ dev and devtest sets in Aragonese.
License
CC-BY-SA-4.0
Attribution
@inproceedings{wmt24-spain,
title="Expanding the FLORES+ Multilingual Benchmark with Translations for {Aragonese, Aranese, Asturian, and Valencian}",
author="Juan Antonio Perez-Ortiz and Felipe S{\'a}nchez-Martínez and Víctor M. S{\'a}nchez-Cartagena and Miquel Esplà-Gomis and Aaron Galiano Jimenez and Antoni Oliver and Claudi Aventín-Boya and Alejandro Pardos and Cristina Valdés and Jus{\'e}p Loís Sans Socasau and Juan Pablo Martínez",
booktitle = "Proceedings of the Ninth Conference on Machine Translation",
month = nov,
year = "2024",
address = "Miami, USA",
publisher = "Association for Computational Linguistics"
}
With the support of the R+D+i projects PID2021-127999NB-I00 (LiLowLa: Lightweight neural translation technologies for low-resource languages) and PID2021-124663OB-I00 (TAN-IBE: Neural Machine Translation for the Romance languages of the Iberian Peninsula), funded by MCIN /AEI /10.13039/501100011033 / FEDER, UE.
Language codes
- ISO-639-3: arg
- ISO 15924: Latn
- Glottocode: arag1245
Additional language information
- Academia de l'Aragonés
- Orthography of Aragonese, published by EFA-ACAR (Estudio de Filología Aragonesa – Academia de l’Aragonés).
- Grammar of Aragonese, published by EFA-ACAR.
- Dictionary of Aragonese published by EFA-ACAR.
- Aragonese dictionary Tresoro d'a Luenga Aragonesa.
- Spanish--Aragonese and Aragonese--Spanish dictionary Aragonario, developed under the project LINGUATEC and published by the Government of Aragon.
- Spanish-Aragonese machine translation system
Workflow
The data for Aragonese was initially translated from Spanish using the Spanish-Aragonese Apertium rule-based machine translation system. This translation was subsequently post-edited by specialists proficient in Aragonese. The utilization of machine translation is justified for two reasons: firstly, the two-step workflow consisting of machine translation followed by post-editing is prevalent for this language, with many existing texts being produced this way; secondly, sourcing linguists or translators for Aragonese proved challenging due to their scarcity, making it difficult to complete the task within the required timeframe.
Finally, the post-edited translation was reviewed by a member of the Academia Aragonesa de la Lengua. In this last step, the reviewer had also in view the English version of the FLORES+ dataset; an important number of translations were modified for better agreement with English original data.
Additional guidelines
The guidelines provided by the Academia de l'Aragonés were followed to ensure that the translation into Aragonese aligned with their recommendations.