Lune

ACL Findings2025

Low-Resource Grammatical Error Correction: Selective Data Augmentation with Round-Trip Machine Translation

Frank Palma Gomez, Alla Rozovskaya

Abstract

Supervised state-of-the-art methods for grammatical error correction require large amounts of parallel data for training. Due to lack of goldlabeled data, techniques that create synthetic training data have become popular. We show that models trained on synthetic data tend to correct a limited range of grammar and spelling mistakes that involve character-level changes, but perform poorly on (more complex) phenomena that require word-level changes. We propose to address the performance gap on such errors by generating synthetic data through selective data augmentation via round-trip machine translation. We show that the proposed technique, SeLex-RT, is capable of generating mistakes that are similar to those observed with language learners. Using the approach with two types of state-of-the-art learning frameworks and two low-resource languages (Russian and Ukrainian), we achieve substantial improvements, compared to training on synthetic data produced with standard techniques. Analysis of the output reveals that models trained on data noisified with the SeLex-RT approach are capable of making word-level changes and correct lexical errors common with language learners. 1

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

Questions to start from

Lune can draw on

the full paper · related work

Get the answer

Free to start. No credit card required.

Lineage

  • MixEdit: Revisiting Data Augmentation and Beyond for Grammatical Error Correction cites LM-Critic: Language Models for Unsupervised Grammatical Error Correction
  • MixEdit: Revisiting Data Augmentation and Beyond for Grammatical Error Correction cites Improving Grammatical Error Correction with Machine Translation Pairs
Builds onCited byLeft to right by year; the more cited, the higher.

Builds on6

  1. Improving Grammatical Error Correction with Machine Translation PairsWangchunshu Zhou, Tao Ge, Chang Mu, Ke Xu et al.EMNLP Findings 2020 · 44 citations
  2. LM-Critic: Language Models for Unsupervised Grammatical Error CorrectionMichihiro Yasunaga, Jure Leskovec, Percy LiangEMNLP 2021 · 29 citations
  3. Type-Driven Multi-Turn Corrections for Grammatical Error CorrectionShaopeng Lai, Qingyu Zhou, Jiali Zeng, Zhongli Li et al.ACL Findings 2022 · 20 citations
  4. Advancements in Arabic Grammatical Error Detection and Correction: An Empirical InvestigationBashar Alhafni, Go Inoue, Christian Khairallah, Nizar HabashEMNLP 2023 · 10 citations
  5. MixEdit: Revisiting Data Augmentation and Beyond for Grammatical Error CorrectionJingheng Ye, Yinghui Li, Yangning Li, Hai-Tao ZhengEMNLP Findings 2023 · 6 citations

Related papers

  1. Improving Grammatical Error Correction via Contextual Data AugmentationYixuan Wang, Baoxin Wang, Yijun Liu, Qingfu Zhu et al.ACL Findings 2024 · 6 citations
  2. TransGEC: Improving Grammatical Error Correction with TranslationeseTao Fang, Xuebo Liu, Derek F. Wong, Runzhe Zhan et al.ACL Findings 2023 · 11 citations
  3. DAGA: Data Augmentation with a Generation Approach forLow-resource Tagging TasksBosheng Ding, Linlin Liu, Lidong Bing, Canasai Kruengkrai et al.EMNLP 2020 · 132 citations
  4. NSL-MT: Linguistically Informed Negative Samples for Efficient Machine Translation in African Low-Resource LanguagesMamadou K. Keita, Christopher M. Homan, Huy LeACL Findings 2026 · 2 citations
  5. Optimized Tokenization for Transcribed Error CorrectionTomer Wullach, Shlomo E. ChazanEMNLP 2023 · 1 citation
Dusk over the sea between two cliffs drawn in fine vertical lines