# Reaching Human-level Performance in Automatic Grammatical Error Correction: An Empirical Study

Source: [Reaching Human-level Performance in Automatic Grammatical Error Correction: An Empirical Study](<https://www.sota2.com/research/paper/68fa649c1e209131cea058c5>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T03:00:17.586Z

Authors: Tao Ge, Furu Wei, Ming Zhou

Published: 2018-07-03T16:37:05.000Z

## Abstract

Neural sequence-to-sequence (seq2seq) approaches have proven to be successful in grammatical error correction (GEC). Based on the seq2seq framework, we propose a novel fluency boost learning and inference mechanism. Fluency boosting learning generates diverse error-corrected sentence pairs during training, enabling the error correction model to learn how to improve a sentence's fluency from more instances, while fluency boosting inference allows the model to correct a sentence incrementally with multiple inference steps. Combining fluency boost learning and inference with convolutional seq2seq models, our approach achieves the state-of-the-art performance: 75.72 (F\_{0.5}) on CoNLL-2014 10 annotation dataset and 62.42 (GLEU) on JFLEG test set respectively, becoming the first GEC system that reaches human-level performance (72.58 for CoNLL and 62.37 for JFLEG) on both of the benchmarks.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Grammatical Error Correction on CoNLL-10](<https://www.sota2.com/research/sota/grammatical-error-correction-on-conll-10.md>): Grammatical Error Correction; CoNLL-10; F0.5 Score: 76.88.

- [Grammatical Error Correction on CoNLL-10 SvH](<https://www.sota2.com/research/sota/grammatical-error-correction-on-conll-10-svh.md>): Grammatical Error Correction; CoNLL-10 SvH; F0.5 Score: 75.93.

- [Grammatical Error Correction on JFLEG](<https://www.sota2.com/research/sota/grammatical-error-correction-on-jfleg.md>): Grammatical Error Correction; JFLEG; GLEU: 62.42.

- [Grammatical Error Correction on JFLEG (test)](<https://www.sota2.com/research/sota/grammatical-error-correction-on-jfleg-test.md>): Grammatical Error Correction; JFLEG (test); GLEU: 62.42.

- [Grammatical Error Correction on CoNLL 2014](<https://www.sota2.com/research/sota/grammatical-error-correction-on-conll-2014.md>): Grammatical Error Correction; CoNLL 2014; F0.5: 61.34.

- [Grammatical Error Correction on CoNLL 2014 (test)](<https://www.sota2.com/research/sota/grammatical-error-correction-on-conll-2014-test.md>): Grammatical Error Correction; CoNLL 2014 (test); F0.5 Score: 61.34.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/1807.01270>)

[PDF](<https://arxiv.org/pdf/1807.01270.pdf>)

[Implementation](<https://github.com/getao/human-performance-gec>)
