Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2511.09537 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917462410264576 |
|---|---|
| author | Keita, Mamadou K. Homan, Christopher Le, Huy |
| author_facet | Keita, Mamadou K. Homan, Christopher Le, Huy |
| contents | We introduce negative space learning machine translation (NSL-MT), a training method for underresourced languages, that augments limited parallel data with synthetically generated violations of the target language's grammar and explicitly penalizes the model when it assigns high probability to these linguistically invalid outputs. NSL-MT delivers improvements across all baselines we tested, including 3-12% BLEU gains for well-performing models and 56-89% gains for models lacking decent initial support. Furthermore, NSL-MT provides a 5x data efficiency multiplier: training with 1,000 examples matches or exceeds normal training with 5,000 examples. NSL-MT thus provides a data-efficient alternative training method for settings where parallel data is limited. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_09537 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | NSL-MT: Linguistically Informed Negative Samples for Efficient Machine Translation in Low-Resource Languages Keita, Mamadou K. Homan, Christopher Le, Huy Machine Learning We introduce negative space learning machine translation (NSL-MT), a training method for underresourced languages, that augments limited parallel data with synthetically generated violations of the target language's grammar and explicitly penalizes the model when it assigns high probability to these linguistically invalid outputs. NSL-MT delivers improvements across all baselines we tested, including 3-12% BLEU gains for well-performing models and 56-89% gains for models lacking decent initial support. Furthermore, NSL-MT provides a 5x data efficiency multiplier: training with 1,000 examples matches or exceeds normal training with 5,000 examples. NSL-MT thus provides a data-efficient alternative training method for settings where parallel data is limited. |
| title | NSL-MT: Linguistically Informed Negative Samples for Efficient Machine Translation in Low-Resource Languages |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2511.09537 |