| Language | Rank | Precision | Recall | F1 |
|---|---|---|---|---|
| Kannada | 1 | 0.73 | 0.78 | 0.75 |
| Malayalam | 2 | 0.97 | 0.97 | 0.96 |
| Tamil | 3 | 0.75 | 0.79 | 0.76 |
- See ./code/datasets/eacl2021 for creating data and running code. However, the main scripts are located in code/scripts folder.
- The .ipynb files (colab files) use absolute paths. So be careful and please change them according to your local directory path(s).
- See ./code/datasets/eacl2021/run.sh for how to run the code. Alternatively, see notebooks in ./colab_notebooks
- Pretrained models, submission files and training checkpoints can be downloaded from this drive repo.
- Scripts for task-adaptive pretraining are placed at ./pretraining
@misc{jayanthi2021sjajdravidianlangtecheacl2021,
title={SJ_AJ@DravidianLangTech-EACL2021: Task-Adaptive Pre-Training of Multilingual BERT models for Offensive Language Identification},
author={Sai Muralidhar Jayanthi and Akshat Gupta},
year={2021},
eprint={2102.01051},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
Feel free to contact us for a quick chat or a discussion at gmail