Official code for the ACL 2022 paper: Can Unsupervised Knowledge Transfer from Social Discussions Help Argument Mining?.
To run experiments, copy the corresponding file from exp_scripts/ into the folder containing this repo and run the script from there.
See README.md files within for further instructions.
| Model Name | Description | Link |
|---|---|---|
| sMLM-LF | Longformer Pretrained using selective-MLM with local attention window size 512 | sMLM-LF |
| sMLM-256-LF | Longformer Pretrained using selective-MLM with local attention window size 256 | sMLM-256-LF |
| DA-LF | Longformer domain adapted to Reddit CMV Modes data | DA-LF |
| sMLM-RoBERTa | RoBERTa trained with selective-MLM | sMLM-RoBERTa |
| DA-BERT | BERT domain adapted to Reddit CMV Modes Data | DA-BERT |