Skip to main content
Språkbanken Text is a part of Språkbanken.

MultiGED

Standard reference Information

Elena Volodina, Christopher Bryant, Andrew Caines, Orphée De Clercq, Jennifer-Carmen Frey, Elizaveta Ershova, Alexandr Rosen, Olga Vinogradova (2023): MultiGED-2023 shared task at NLP4CALL: Multilingual Grammatical Error Detection, in Proceedings of the 12th Workshop on Natural Language Processing for Computer Assisted Language Learning (NLP4CALL 2023) BibTeX

Data citation Information

Elena Volodina, Chris Bryant, Andrew Caines, Orphée De Clercq, Jennifer-Carmen Frey, Elizaveta Ershova, Alexandr Rosen, & Olga Vinogradova (2025). MultiGED (updated: 2025-01-19). [Data set]. Språkbanken Text. https://doi.org/10.23695/xe7r-k506
BibTeX Additional ways to cite the dataset.
MultiGEC is a dataset for Grammatical Error Detection (a task within NLP) containing data for 5 languages (Czech, English, German, Italian and Swedish).


Dataset description

MultiGED is a dataset for Multilingual Grammatical Error Detection in 5 European languages (Czech, English, German, Italian and Swedish) compiled by the CompSLA working group in the context of MultiGED-2023, the first multilinual GED shared task.

The data comes from learner essays, but the sequence of sentences within essays is not kept. Instead, this is a set of randomized sentences to prevent re-construction of original essays.

Data is provided in a tab-separated format consisting of two columns, where the first column contains the token and the second column contains the label (c or i), i.e. correct and incorrect. Note that there are no column headers, each sentence is separated by an empty line, and double quotes are escaped. See more on data format .

Annotation

Each token has a label c or i - for correct and incorrect.

Caveats

The data is relatively homogeneous, although the definition of an error span is slightly different between the included languages.

Intended uses

Grammatical Error Detection, (Second) Language Acquisiton studies, Learner Corpus Research, Noisy User-produced Data.

References

Accessible through

Access Platform Licence
varies
varies

Download

File Size Modified Licence
3.82 MB 2025-01-22

Type

  • Corpus
  • Training and evaluation data

Language

Czech
German
English
Italian
Swedish

Size

Keywords

  • grammatical error detection
  • token-level detection
  • language learning
  • sentences
  • multilinguality

Creators

  • Elena Volodina
  • Chris Bryant
  • Andrew Caines
  • Orphée De Clercq
  • Jennifer-Carmen Frey
  • Elizaveta Ershova
  • Alexandr Rosen
  • Olga Vinogradova

Updated

2025-01-19

Contact

Språkbanken Text, Sweden
sb-info@svenska.gu.se