These metrics were used in Courtney Napoles, Keisuke Sakaguchi, and Joel Tetreault. There's No Comparison: Reference-less Evaluation Metrics in Grammatical Error Correction. EMNLP 2016
If you use this code or the accompanying CodaLab evaluation, please cite:
@InProceedings{napoles-sakaguchi-tetreault:2016:EMNLP2016,
author = {Napoles, Courtney and Sakaguchi, Keisuke and Tetreault, Joel},
title = {There's No Comparison: Reference-less Evaluation Metrics in Grammatical Error Correction},
booktitle = {Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing},
month = {November},
year = {2016},
address = {Austin, Texas},
publisher = {Association for Computational Linguistics},
pages = {2109--2115},
url = {https://aclweb.org/anthology/D16-1228}
}
The CodaLab evaluation can be used for to evaluate grammatical error correcions of the CoNLL-2014 shared task test set.
https://competitions.codalab.org/competitions/15475
codalab/
heilman-et-al/