© [2024] Ubisoft Entertainment. All Rights Reserved
Synthetically corrupted sentence pairs for grammatical error detection.
Each item contains a correct sentence and an incorrect (corrupted) version
generated by an NLLB-based corruption model.
| Code | Language | Items |
|---|---|---|
| ar | Arabic | 295,934 |
| cs | Czech | 301,059 |
| de | German | 363,818 |
| en | English | 385,204 |
| es | Spanish | 364,752 |
| et | Estonian | 135,760 |
| is | Icelandic | 209,126 |
| it | Italian | 348,338 |
| ru | Russian | 441,220 |
| sv | Swedish | 385,373 |
| zh | Chinese | 390,448 |
| - | Total | 3,621,032 |
for EMNLP 24 - Zero-shot Cross-Lingual Transfer for Synthetic Data Generation in Grammatical Error Detection ( https://arxiv.org/pdf/2407.11854 )
© [2024] Ubisoft Entertainment. All Rights Reserved