# Bridging the Data Gap between Training and Inference for Unsupervised Neural Machine Translation

arXiv: https://arxiv.org/abs/2203.08394 · HTML: https://mlsanity.com/paper/2203.08394/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 37 references that are in the map.
- First posted on arXiv 16 Mar 2022.

## Builds on (most cited first, 23 of 23)

- [A Call for Clarity in Reporting BLEU Scores](https://mlsanity.com/paper/1804.08771/) (2018)
- [Cross-lingual Language Model Pretraining](https://mlsanity.com/paper/1901.07291/) (2019)
- [Improving Neural Machine Translation Models with Monolingual Data](https://mlsanity.com/paper/1511.06709/) (2015)
- [Word Translation Without Parallel Data](https://mlsanity.com/paper/1710.04087/) (2017)
- [MASS: Masked Sequence to Sequence Pre-training for Language Generation](https://mlsanity.com/paper/1905.02450/) (2019)
- [Unsupervised Machine Translation Using Monolingual Corpora Only](https://mlsanity.com/paper/1711.00043/) (2017)
- [Revisiting Self-Training for Neural Sequence Generation](https://mlsanity.com/paper/1909.13788/) (2019)
- [Unsupervised Neural Machine Translation](https://mlsanity.com/paper/1710.11041/) (2017)
- [A robust self-learning method for fully unsupervised cross-lingual mappings of word embeddings](https://mlsanity.com/paper/1805.06297/) (2018)
- [Phrase-Based & Neural Unsupervised Machine Translation](https://mlsanity.com/paper/1804.07755/) (2018)
- [On Exposure Bias, Hallucination and Domain Shift in Neural Machine Translation](https://mlsanity.com/paper/2005.03642/) (2020)
- [BLEU might be Guilty but References are not Innocent](https://mlsanity.com/paper/2004.06063/) (2020)
- [Cross-lingual Retrieval for Iterative Self-Supervised Training](https://mlsanity.com/paper/2006.09526/) (2020)
- [The Effect of Translationese in Machine Translation Test Sets](https://mlsanity.com/paper/1906.08069/) (2019)
- [On The Evaluation of Machine Translation Systems Trained With Back-Translation](https://mlsanity.com/paper/1908.05204/) (2019)
- [An Effective Approach to Unsupervised Machine Translation](https://mlsanity.com/paper/1902.01313/) (2019)
- [Unsupervised Neural Machine Translation with Generative Language Models Only](https://mlsanity.com/paper/2110.05448/) (2021)
- [Self-Training Sampling with Monolingual Data Uncertainty for Neural Machine Translation](https://mlsanity.com/paper/2106.00941/) (2021)
- [Unsupervised Neural Machine Translation with Weight Sharing](https://mlsanity.com/paper/1804.09057/) (2018)
- [On the Language Coverage Bias for Neural Machine Translation](https://mlsanity.com/paper/2106.03297/) (2021)
- [Unsupervised Neural Machine Translation with SMT as Posterior Regularization](https://mlsanity.com/paper/1901.04112/) (2019)
- [Self-Training for Unsupervised Neural Machine Translation in Unbalanced Training Data Scenarios](https://mlsanity.com/paper/2004.04507/) (2020)
- [Cross-model Back-translated Distillation for Unsupervised Machine Translation](https://mlsanity.com/paper/2006.02163/) (2020)

## Built on by (most cited first, 5 of 5)

- [Exploring Human-Like Translation Strategy with Large Language Models](https://mlsanity.com/paper/2305.04118/) (2023)
- [Tencent AI Lab - Shanghai Jiao Tong University Low-Resource Translation System for the WMT22 Translation Task](https://mlsanity.com/paper/2210.08742/) (2022)
- [The first neural machine translation system for the Erzya language](https://mlsanity.com/paper/2209.09368/) (2022)
- [A Theory of Unsupervised Translation Motivated by Understanding Animal Communication](https://mlsanity.com/paper/2211.11081/) (2022)
- [Model-Based Offline Reinforcement Learning with Reliability-Guaranteed Sequence Modeling](https://mlsanity.com/paper/2502.06491/) (2025)

## Abstract (arXiv)

Back-translation is a critical component of Unsupervised Neural Machine Translation (UNMT), which generates pseudo parallel data from target monolingual data. A UNMT model is trained on the pseudo parallel data with translated source, and translates natural source sentences in inference. The source discrepancy between training and inference hinders the translation performance of UNMT models. By carefully designing experiments, we identify two representative characteristics of the data gap in source: (1) style gap (i.e., translated vs. natural text style) that leads to poor generalization capability; (2) content gap that induces the model to produce hallucination content biased towards the target language. To narrow the data gap, we propose an online self-training approach, which simultaneously uses the pseudo parallel data {natural source, translated target} to mimic the inference scenario. Experimental results on several widely-used language pairs show that our approach outperforms two strong baselines (XLM and MASS) by remedying the style and content gaps.