Argument Pair Extraction

This repository hosts the code for the paper: Have my arguments been replied to? Argument Pair Extraction as Machine Reading Comprehension.

The original code is available in this repo, but I simplified it for the NLP students at the University of Indonesia to do their final project.

Steps to Run Locally

First, create a Conda environment---resource to install Miniconda---with the following command:

conda create -n ape python=3.10
conda activate ape

Second, install the requirements with the following command:

pip install -r requirements.txt

Third, run the scripts in downloader.ipynb in data/ and longformer-base/ to download data and longformer model.

Fourth, create a folder in data/ named processed/

Last, run the following command:

python prepare.py  
python to_bioes.py  
python run.py

Notice that there will be a new folder named saved_models, which consists of the trained model and a log file detailing the F1 score(s) for each epoch.

Alternative

You may compile everything into a Jupyter Notebook file and run it through platforms such as Kaggle or Colab. By doing this, you can leverage free GPUs to accelerate the process.