Neural GEC Systems with Unsupervised Pre-Training on Synthetic Data

This repository contains models, system configurations and outputs of our winning GEC systems in the BEA 2019 shared task described in R. Grundkiewicz, M. Junczys-Dowmunt, K. Heafield: Neural Grammatical Error Correction Systems with Unsupervised Pre-training on Synthetic Data, BEA 2019.

Citation

@inproceedings{grundkiewicz-etal-2019-neural,
    title = "Neural Grammatical Error Correction Systems with Unsupervised Pre-training on Synthetic Data",
    author = "Grundkiewicz, Roman  and
        Junczys-Dowmunt, Marcin  and
        Heafield, Kenneth",
    booktitle = "Proceedings of the Fourteenth Workshop on Innovative Use of NLP for Building Educational Applications",
    month = aug,
    year = "2019",
    address = "Florence, Italy",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/W19-4427",
    pages = "252--263",
}

Content

• systems - original GEC systems developed for and submitted to the restricted and low-resource tracks
• outputs - corrected output and evaluation scores for common GEC test sets
• training - updated training scripts re-producing our GEC system

See README files in each subdirectory for more information. In case of any questions, please open an issue or send me (Roman) an email.