DFKI SLT at GermEval 2021

Code for the paper: Multilingual Pre-training and Data Augmentation for the Classification of Toxicity in Social Media Comments

Datasets

The data folder contains all the files used to train the models. Some files might not be used for the final versions of the models.

To train the models, please run:

python train.py <parameters>

Run python train.py --help to see all the posibilities.

All the final models are available on the HuggingFace Hub.

Model	Link
GBERT (Fine-tuning only)	https://huggingface.co/airKlizz/gbert-base-germeval21-toxic
XLM-RoBERTa (Fine-tuning only)	https://huggingface.co/airKlizz/xlm-roberta-base-germeval21-toxic
mT5 (Fine-tuning only)	https://huggingface.co/airKlizz/mt5-base-germeval21-toxic
XLM-RoBERTa (With task-specific pre-training)	https://huggingface.co/airKlizz/xlm-roberta-base-germeval21-toxic-with-task-specific-pretraining
mT5 (With task-specific pre-training)	https://huggingface.co/airKlizz/mt5-base-germeval21-toxic-with-task-specific-pretraining
GBERT (With data augmentation)	https://huggingface.co/airKlizz/gbert-base-germeval21-toxic-with-data-augmentation
XLM-RoBERTa (With data augmentation)	https://huggingface.co/airKlizz/xlm-roberta-base-germeval21-toxic-with-data-augmentation
mT5 (With data augmentation)	https://huggingface.co/airKlizz/mt5-base-germeval21-toxic-with-data-augmentation
XLM-RoBERTa (With task-specific pre-training and data augmentation)	https://huggingface.co/airKlizz/xlm-roberta-base-germeval21-toxic-with-task-specific-pretraining-and-data-augmentation
mT5 (With task-specific pre-training and data augmentation)	https://huggingface.co/airKlizz/mt5-base-germeval21-toxic-with-task-specific-pretraining-and-data-augmentation