MT-based Distantly Supervised Grammatical Error Correction for Swedish

This repository contains the source code and the models used for Shared task on Multilingual Grammatical Error Detection (MultiGED-2023). The system utilizes a neural machine translation-based model trained on an automatically corrupted large corpus of Swedish text (~3.2 billion words).

The model is intended as a baseline for our ongoing work on Swedish grammatical error correction using large language models (Östling&Kurfali, 2022). The shared task results suggest that our model is conservative yet remarkably precise in its predictions.

Prerequisites

• Python 3.7 or higher
• SentencePiece
• OpenNMT==2.3.0

Usage

1. Clone the repository: git clone https://github.com/your-username/grammatical-error-detection-swedish.git

2. Download the model from "Google Drive".

3. Run the script to process your input:

bash correct.sh <model_path> <input_file> <output_directory>

• <model_path>: Path to the pre-trained model file.
• <input_file>: Path to the input file to be processed.
• <output_directory>: Path to the directory where the output files will be saved.

The input file should follow the "one sentence per line" format.

Bibtex Citation

If you use this system or find it helpful, please consider citing the accompanying paper:

@inproceedings{kurfali2023distantly,
  title={A distantly supervised Grammatical Error Detection/Correction system for Swedish},
  author={Kurfal{\i}, Murathan and {\"O}stling, Robert},
  booktitle={Swedish Language Technology Conference and NLP4CALL},
  pages={35--39},
  year={2023}
}