Grammatical Error Correction
Grammatical Error Correction (GEC) is the task of correcting different kinds of errors in text such as spelling, punctuation, grammatical, and word choice errors.
GEC is typically formulated as a sentence correction task. A GEC system takes a potentially erroneous sentence as input and is expected to transform it to its corrected version. See the example given below:
| Input (Erroneous) | Output (Corrected) | | ------------------------- | ---------------------- | |She see Tom is catched by policeman in park at last night. | She saw Tom caught by a policeman in the park last night.|
Papers
Showing 51–60 of 415 papers
All datasetsCoNLL-2014 Shared TaskBEA-2019 (test)Falko-MERLINJFLEGUA-GECCoNLL-2014 Shared Task (10 annotations)RestrictedUnrestricted_Restricted_EstGEC-L2FCGECMuCGEC
Benchmark Results
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Ensembles of best 7 models + GRECO + GTP-rerank | F0.5 | 72.8 | — | Unverified |
| 2 | Majority-voting ensemble on best 7 models | F0.5 | 71.8 | — | Unverified |
| 3 | GRECO (voting+ESC) | F0.5 | 71.12 | — | Unverified |
| 4 | GEC-DI (LM+GED) | F0.5 | 69.6 | — | Unverified |
| 5 | Unsupervised GEC + cLang8 | F0.5 | 69.6 | — | Unverified |
| 6 | ESC | F0.5 | 69.51 | — | Unverified |
| 7 | T5 | F0.5 | 68.87 | — | Unverified |
| 8 | MoECE | F0.5 | 67.79 | — | Unverified |
| 9 | SynGEC | F0.5 | 67.6 | — | Unverified |
| 10 | Sequence tagging + token-level transformations + two-stage fine-tuning (+BERT, RoBERTa, XLNet) | F0.5 | 66.5 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Majority-voting ensemble on best 7 models | F0.5 | 81.4 | — | Unverified |
| 2 | GRECO (voting+ESC) | F0.5 | 80.84 | — | Unverified |
| 3 | ESC | F0.5 | 79.9 | — | Unverified |
| 4 | RedPenNet | F0.5 | 77.6 | — | Unverified |
| 5 | clang_large_ft2-gector | F0.5 | 77.1 | — | Unverified |
| 6 | Unsupervised GEC + cLang8 | F0.5 | 76.5 | — | Unverified |
| 7 | DeBERTa + RoBERTa + XLNet | F0.5 | 76.05 | — | Unverified |
| 8 | MoECE | F0.5 | 74.07 | — | Unverified |
| 9 | Sequence tagging + token-level transformations + two-stage fine-tuning (+RoBERTa, XLNet) | F0.5 | 73.7 | — | Unverified |
| 10 | BEA Combination | F0.5 | 73.2 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Llama + 1M BT + gold | F0.5 | 76.75 | — | Unverified |
| 2 | mT5-based multimodal MoE | F0.5 | 76.3 | — | Unverified |
| 3 | gT5 xxl | F0.5 | 75.96 | — | Unverified |
| 4 | Transformer | F0.5 | 73.71 | — | Unverified |
| 5 | Transformer - synthetic pretrain only | F0.5 | 51.41 | — | Unverified |
| 6 | Multilayer Convolutional Encoder-Decoder | F0.5 | 43.35 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | VERNet | GLEU | 62.1 | — | Unverified |
| 2 | Transformer + Pre-train with Pseudo Data + BERT | GLEU | 62 | — | Unverified |
| 3 | SMT + BiGRU | GLEU | 61.5 | — | Unverified |
| 4 | Copy-augmented Model (4 Ensemble +Denoising Autoencoder) | GLEU | 61 | — | Unverified |
| 5 | Transformer | GLEU | 59.9 | — | Unverified |
| 6 | CNN Seq2Seq | GLEU | 57.47 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Llama + 1M BT + gold | F0.5 | 74.09 | — | Unverified |
| 2 | mBART-based model with synthetic data | F0.5 | 68.17 | — | Unverified |
| 3 | mT5 large + 10M synth | F0.5 | 68.09 | — | Unverified |
| 4 | RedPenNet | F0.5 | 67.71 | — | Unverified |
| 5 | ChatGPT (zero-shot) | F0.5 | 27.4 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | GRECO (vote+ESC) | F0.5 | 85.21 | — | Unverified |
| 2 | SMT + BiGRU | F0.5 | 72.04 | — | Unverified |
| 3 | CNN Seq2Seq | F0.5 | 70.14 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | CNN Seq2Seq + Quality Estimation | F0.5 | 56.52 | — | Unverified |
| 2 | Transformer | F0.5 | 55.8 | — | Unverified |
| 3 | + BIFI with no critic | F0.5 | 18.7 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | CNN Seq2Seq + Fluency Boost and inference | GLEU | 62.37 | — | Unverified |
| 2 | CNN Seq2Seq + Fluency Boost | F0.5 | 61.34 | — | Unverified |
| 3 | + BIFI (ours) | F0.5 | 42.4 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Transformer | GLEU | 59.9 | — | Unverified |
| 2 | CNN Seq2Seq | GLEU | 57.47 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Llama + 1M BT + gold | F0.5 | 69.97 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | STG-Joint | exact match | 34.1 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | GEC-DI (LM+GED) | F0.5 | 48.61 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | RedPenNet | F0.5 | 77.6 | — | Unverified |