# Improving Zero-Shot Translation of Low-Resource Languages

arXiv: https://arxiv.org/abs/1811.01389 · HTML: https://mlsanity.com/paper/1811.01389/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 4 Nov 2018.

## Built on by (most cited first, 11 of 11)

- [Isometric MT: Neural Machine Translation for Automatic Dubbing](https://mlsanity.com/paper/2112.08682/) (2021)
- [Prose2Poem: The Blessing of Transformers in Translating Prose to Persian Poetry](https://mlsanity.com/paper/2109.14934/) (2021)
- [Neural Machine Translation for the Indigenous Languages of the Americas: An Introduction](https://mlsanity.com/paper/2306.06804/) (2023)
- [An Empirical Investigation of Multi-bridge Multilingual NMT models](https://mlsanity.com/paper/2110.07304/) (2021)
- [Mergen: The First Manchu-Korean Machine Translation Model Trained on Augmented Data](https://mlsanity.com/paper/2311.17492/) (2023)
- [Self-Learning for Zero Shot Neural Machine Translation](https://mlsanity.com/paper/2103.05951/) (2021)
- [Learning Policies for Multilingual Training of Neural Machine Translation Systems](https://mlsanity.com/paper/2103.06964/) (2021)
- [A Survey of Recent Abstract Summarization Techniques](https://mlsanity.com/paper/2105.00824/) (2021)
- [Multilingual Self-Taught Faithfulness Evaluators](https://mlsanity.com/paper/2507.20752/) (2025)
- [Improving Zero-shot Multilingual Neural Machine Translation for Low-Resource Languages](https://mlsanity.com/paper/2110.00712/) (2021)
- [Exploring Linguistic Similarity and Zero-Shot Learning for Multilingual Translation of Dravidian Languages](https://mlsanity.com/paper/2308.05574/) (2023)

## Abstract (arXiv)

Recent work on multilingual neural machine translation reported competitive performance with respect to bilingual models and surprisingly good performance even on (zeroshot) translation directions not observed at training time. We investigate here a zero-shot translation in a particularly lowresource multilingual setting. We propose a simple iterative training procedure that leverages a duality of translations directly generated by the system for the zero-shot directions. The translations produced by the system (sub-optimal since they contain mixed language from the shared vocabulary), are then used together with the original parallel data to feed and iteratively re-train the multilingual network. Over time, this allows the system to learn from its own generated and increasingly better output. Our approach shows to be effective in improving the two zero-shot directions of our multilingual model. In particular, we observed gains of about 9 BLEU points over a baseline multilingual model and up to 2.08 BLEU over a pivoting mechanism using two bilingual models. Further analysis shows that there is also a slight improvement in the non-zero-shot language directions.