DeMuX: Data-efficient Multilingual Learning

Simran Khanuja, Srinivas Gowriraj, Lucio Dery, Graham Neubig


Abstract
Pre-trained multilingual models have enabled deployment of NLP technologies for multiple languages. However, optimally fine-tuning these models under an annotation budget, such that performance on desired target languages is jointly maximized, still remains an open question. In this paper, we introduce DeMuX, a framework that prescribes the exact data-points to label from vast amounts of unlabelled multilingual data, having unknown degrees of overlap with the target set. Unlike most prior works, our end-to-end framework is language-agnostic, accounts for model representations, and supports multilingual target configurations. Our active learning strategies rely upon distance and uncertainty measures to select task-specific neighbors that are most informative to label, given a model. DeMuX outperforms strong baselines in 84% of the test cases, in the zero-shot setting of disjoint source and target language sets (including multilingual target pools), across three models and four tasks. Notably, in low-budget settings (5-100 examples), we observe gains of up to 8-11 F1 points. Our code is released here: https://github.com/simran-khanuja/demux.
Anthology ID:
2024.naacl-long.412
Volume:
Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
Month:
June
Year:
2024
Address:
Mexico City, Mexico
Editors:
Kevin Duh, Helena Gomez, Steven Bethard
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7423–7436
Language:
URL:
https://aclanthology.org/2024.naacl-long.412
DOI:
10.18653/v1/2024.naacl-long.412
Bibkey:
Cite (ACL):
Simran Khanuja, Srinivas Gowriraj, Lucio Dery, and Graham Neubig. 2024. DeMuX: Data-efficient Multilingual Learning. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 7423–7436, Mexico City, Mexico. Association for Computational Linguistics.
Cite (Informal):
DeMuX: Data-efficient Multilingual Learning (Khanuja et al., NAACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.naacl-long.412.pdf