mlsanityData to 7 Oct 2026

Training on Synthetic Noise Improves Robustness to Natural Noise in Machine Translation

arXiv 1902.01509 ↗First posted 5 Feb 2019cs.CL cs.LG stat.MLcited by 52

Not on the map. Why?

Abstract (arXiv)

We consider the problem of making machine translation more robust to character-level variation at the source side, such as typos. Existing methods achieve greater coverage by applying subword models such as byte-pair encoding (BPE) and character-level encoders, but these methods are highly sensitive to spelling mistakes. We show how training on a mild amount of random synthetic noise can dramatically improve robustness to these variations, without diminishing performance on clean text. We focus on translation performance on natural noise, as captured by frequent corrections in Wikipedia edit logs, and show that robustness to such noise can be achieved using a balanced diet of simple synthetic noises at training time, without access to the natural noise data or distribution.

Key facts. Cited by 52 papers in the map (data through 7 Oct 2026). First posted on arXiv 5 Feb 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 12; 2022: 9; 2023: 14; 2024: 11; 2025: 4; 2026: 2.

52 papers.

Sort:

By area:

  1. 2023Hallucinations in Large Multilingual Translation Modelscited by 126
  2. 2021Measure and Improve Robustness in NLP Models: A Surveycited by 107
  3. 2022Why Should Adversarial Perturbations be Imperceptible? Rethink the Research Paradigm in Adversarial NLPcited by 99
  4. 2021Domain Adaptation and Multi-Domain Adaptation for Neural Machine Translation: A Surveycited by 46
  5. 2023Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspectioncited by 44
  6. 2022Faithfulness in Natural Language Generation: A Systematic Survey of Analysis, Evaluation and Optimization Methodscited by 27
  7. 2021Improving Zero-shot Cross-lingual Transfer between Closely Related Languages by injecting Character-level Noisecited by 19
  8. 2021To Augment or Not to Augment? A Comparative Study on Text Augmentation Techniques for Low-Resource NLPcited by 12
  9. 2021On the Robustness of Intent Classification and Slot Labeling in Goal-oriented Dialog Systems to Real-world Noisecited by 9
  10. 2022CipherDAug: Ciphertext based Data Augmentation for Neural Machine Translationcited by 9
  11. 2022Robustification of Multilingual Language Models to Real-world Noise in Crosslingual Zero-shot Settings with Robust Contrastive Pretrainingcited by 9
  12. 2024Did Translation Models Get More Robust Without Anyone Even Noticing?cited by 9
38 more
  1. 2021Empirical Error Modeling Improves Robustness of Noisy Neural Sequence Labelingcited by 8
  2. 2024Improving Black-box Robustness with In-Context Rewritingcited by 8
  3. 2021Addressing the Vulnerability of NMT in Input Perturbationscited by 6
  4. 2023BERTwich: Extending BERT's Capabilities to Model Dialectal and Noisy Textcited by 6
  5. 2024Synthetic Data Generation and Joint Learning for Robust Code-Mixed Translationcited by 5
  6. 2024Stochastic Monkeys at Play: Random Augmentations Cheaply Break LLM Safety Alignmentcited by 5
  7. 2022Generating Authentic Adversarial Examples beyond Meaning-preserving with Doubly Round-trip Translationcited by 4
  8. 2022Improving the Robustness of Summarization Models by Detecting and Removing Input Noisecited by 4
  9. 2021Robust Learning for Text Classification with Multi-source Noise Simulation and Hard Example Miningcited by 3
  10. 2023From Adversarial Arms Race to Model-centric Evaluation: Motivating a Unified Automatic Robustness Evaluation Frameworkcited by 3
  11. 2023ViPE: Visualise Pretty-much Everythingcited by 3
  12. 2023Ask Language Model to Clean Your Noisy Translation Datacited by 3
  13. 2024Semantics or spelling? Probing contextual word embeddings with orthographic noisecited by 3
  14. 2025Misspellings in Natural Language Processing: A surveycited by 3
  15. 2025Sarang at DEFACTIFY 4.0: Detecting AI-Generated Text Using Noised Data and an Ensemble of DeBERTa Modelscited by 3
  16. 2021Visual Cues and Error Correction for Translation Robustnesscited by 2
  17. 2021Secoco: Self-Correcting Encoding for Neural Machine Translationcited by 2
  18. 2022PAEG: Phrase-level Adversarial Example Generation for Neural Machine Translationcited by 2
  19. 2022Reproduction and Replication of an Adversarial Stylometry Experimentcited by 2
  20. 2023Data Augmentation for Neural NLPcited by 2
  21. 2023LEA: Improving Sentence Similarity Robustness to Typos Using Lexical Attention Biascited by 2
  22. 2023Impact of Visual Context on Noisy Multimodal NMT: An Empirical Study for English to Indian Languagescited by 2
  23. 2024Specification Overfitting in Artificial Intelligencecited by 2
  24. 2024CERT-ED: Certifiably Robust Text Classification for Edit Distancecited by 2
  25. 2021Doubly-Trained Adversarial Data Augmentation for Neural Machine Translationcited by 1
  26. 2023CharSpan: Utilizing Lexical Similarity to Enable Zero-Shot Machine Translation for Extremely Low-resource Languagescited by 1
  27. 2023The Paradox of Noise: An Empirical Study of Noise-Infusion Mechanisms to Improve Generalization, Stability, and Privacy in Federated Learningcited by 1
  28. 2024Investigating Training Strategies and Model Robustness of Low-Rank Adaptation for Language Modeling in Speech Recognitioncited by 1
  29. 2024Enhancing Effectiveness and Robustness in a Low-Resource Regime via Decision-Boundary-aware Data Augmentationcited by 1
  30. 2025A Representation Level Analysis of NMT Model Robustness to Grammatical Errorscited by 1
  31. 2021Masked Adversarial Generation for Neural Machine Translationcited by 0
  32. 2022Machine Translation Robustness to Natural Asemantic Variationcited by 0
  33. 2023Optimized Tokenization for Transcribed Error Correctioncited by 0
  34. 2023Is Robustness Transferable across Languages in Multilingual Neural Machine Translation?cited by 0
  35. 2023Evaluating Robustness of Dialogue Summarization Models in the Presence of Naturally Occurring Variationscited by 0
  36. 2024Making Sentence Embeddings Robust to User-Generated Contentcited by 0
  37. 2024An Empirical Study on the Robustness of Massively Multilingual Neural Machine Translationcited by 0
  38. 2025Extend Adversarial Policy Against Neural Machine Translation via Unknown Tokencited by 0

50 most cited of 52 All 52 (JSON)