Which Side Are You On? A Multi-task Dataset for End-to-End Argument Summarisation and Evaluation (ACL2024)

This work followed the paper publised on ACL 2024

Ethic approved

This work has received ethical review from the institution under the license number 2024-16577-36845

Due to the suggestion, we will consider release task 1 and task 2's data in a more safety ways. It will be release very soon!

Requirements

• python
• pytorch
• sentence-transformers

Run Baselines

Task1

Task2

• Classification models

XXX

• Contrastive Learning models

# roberta-large
python train.py --model_name roberta-large \
  --train_dataset_file PATH/TO/train_convincingness.csv \
  --dev_dataset_file PATH/TO/dev_convincingness.csv \
  --test_dataset_file PATH/TO/test_convincingness.csv \
  --output_path PATH/TO/OUTPUT \
  --num_epochs  10 \
  --train_batch_size 32 \
  --eval_batch_size 64 \
  --max_input_length 256 \
  --add_special_tokens "<SEP>" \
  --learning_rate 3e-5 \
  --task_name 'task2'

# sentence-t5-large or sentence-t5-xl
 python train.py --model_name sentence-transformers/sentence-t5-large \
  --train_dataset_file PATH/TO/train_convincingness.csv \
  --dev_dataset_file PATH/TO/dev_convincingness.csv \
  --test_dataset_file PATH/TO/test_convincingness.csv \
  --output_path PATH/TO/OUTPUT \
  --num_epochs  10 \
  --train_batch_size 32 \
  --eval_batch_size 64 \
  --max_input_length 256 \
  --add_special_tokens "</s>" \
  --learning_rate 3e-5 \
  --sentence_transformer \
  --task_name 'task2'

Task3

# roberta-large
python train.py --model_name roberta-large \
  --train_dataset_file PATH/TO/task3_trainset.csv \
  --dev_dataset_file PATH/TO/task3_devset.csv \
  --test_dataset_file PATH/TO/task3_testset.csv \
  --output_path PATH/TO/OUTPUT \
  --num_epochs  10 \
  --train_batch_size 32 \
  --eval_batch_size 64 \
  --max_input_length 512 \
  --learning_rate 3e-5 \
  --task_name 'task3'

# sentence-t5-large or sentence-t5-xl
 python train.py --model_name sentence-transformers/sentence-t5-large \
  --train_dataset_file PATH/TO/task3_trainset.csv \
  --dev_dataset_file PATH/TO/task3_devset.csv \
  --test_dataset_file PATH/TO/task3_testset.csv \
  --output_path PATH/TO/OUTPUT \
  --num_epochs  10 \
  --train_batch_size 32 \
  --eval_batch_size 64 \
  --max_input_length 512 \
  --learning_rate 3e-5 \
  --sentence_transformer \
  --task_name 'task3'

Task4