Minimally-Augmented Grammatical Error Correction

This repository contains models and training scripts for our MAGEC systems for English, German and Russian described in R. Grundkiewicz, M. Junczys-Dowmunt, Minimally-Augmented Grammatical Error Correction, W-NUT 2019.

Citation

@inproceedings{grundkiewicz-etal-2019-neural,
    title = "Minimally-Augmented Grammatical Error Correction",
    author = "Grundkiewicz, Roman and Junczys-Dowmunt, Marcin",
    booktitle = "Proceedings of the 2019 {EMNLP} Workshop W-{NUT}: The 5th Workshop on Noisy User-generated Text",
    year = "2019",
    address = "Hong Kong",
    publisher = "Association for Computational Linguistics",
}

Content

• models - MAGEC models used in the paper for English, German and Russian experiments
• data - training data and scripts for generating synthetic data from monolingual texts
• training - training scripts for MAGEC systems for English, German and Russian

See README files in each subdirectory for more information. In case of any questions, please feel free to contact me (Roman) via email.