Paper @ Neural Computing and Applications and Dataset @ HuggingFace
git clone https://github.com/mehedihasanbijoy/BanglaGEC.git
or manually download and extract the github repository of BanglaGEC.
conda env create -f environment.yml
conda activate BanglaGEC
python -c "from datasets import load_dataset; load_dataset('nahid-hub/BanglaGEC').save_to_disk('./Dataset')"
or manually download the corpus from here and keep the extracted files into ./Dataset/
python main.py --CORPUS_PATH "./Dataset/corpus.csv" --KNOWLEDGE_PATH "./KnowledgeToBeTransferred/paraphrase.pth" --CHECKPOINT_PATH "./Checkpoints/panini.pth" --MODEL_NAME "panini" --BATCH_SIZE 16 --N_EPOCHS 50
python main.py --CORPUS_PATH "./Dataset/corpus.csv" --CHECKPOINT_PATH "./Checkpoints/banglat5.pth" --MODEL_NAME "banglat5" --BATCH_SIZE 16 --N_EPOCHS 50
python main.py --CORPUS_PATH "./Dataset/corpus.csv" --CHECKPOINT_PATH "./Checkpoints/t5small.pth" --MODEL_NAME "t5small" --BATCH_SIZE 16 --N_EPOCHS 50
@article{hossain2024panini,
title={Panini: a transformer-based grammatical error correction method for bangla},
author={Hossain, Nahid and Bijoy, Mehedi Hasan and Islam, Salekul and Shatabda, Swakkhar},
journal={Neural Computing and Applications},
volume={36},
number={7},
pages={3463--3477},
year={2024},
publisher={Springer}
}