Harnessing Toulmin’s theory for zero-shot argument explication

This is replication code to support the paper Harnessing Toulmin’s theory for zero-shot argument explication which is forthcoming in Proceedings of ACL 2024.

If you use this code or data, please cite our paper:

@article{gupta2024harnessing,
  author    = {Gupta, Ankita and Zuckerman, Ethan, and O'Connor, Brendan},
  title     = {Harnessing Toulmin’s theory for zero-shot argument explication},
  journal   = {ACL},
  year      = {2024},
  note      = {Forthcoming}
}

Set-up

Follow these instructions to set-up your repository. You will need to download Anaconda to run conda and python commands.

git clone https://github.com/slanglab/argument_explication.git
cd argument_explication
conda create -y --name argex python==3.9
conda activate argex
pip install -r requirements.txt

Evaluation Datasets

The prior argumentation datasets for evaluation should be placed in the data folder from the following sources:

1. Implicit Knowledge in Argumentative Texts: An annotated Corpus

2. An annotated corpus of argumentative microtexts.

3. The Argument Reasoning Comprehension Task: Identification and Reconstruction of Implicit Warrants