Disarm

Disarm is a RoBERTa based neural system for Argument Relation Classification (ARC). The model leverages discourse markers through adversarial training on the Discovery dataset. In particular, Disarm maps into a joint embedding space the argumentative units from the ARC task and sentences from the Discourse Marker Detection task (DMD). This result is achieved through the use of a Gradient Reversal Layer.

This repository holds the code of the paper "Argument Relation Classification through Discourse Markers and Adversarial Training" published at EMNLP 2024. For a detailed description of the work please read our paper. Please cite the paper if you use the code from this repository in your work!

@inproceedings{contalbo-etal-2024-argument,
    title = "Argument Relation Classification through Discourse Markers and Adversarial Training",
    author = "Contalbo, Michele Luca  and
      Guerra, Francesco  and
      Paganelli, Matteo",
    editor = "Al-Onaizan, Yaser  and
      Bansal, Mohit  and
      Chen, Yun-Nung",
    booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2024",
    address = "Miami, Florida, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.emnlp-main.1054",
    doi = "10.18653/v1/2024.emnlp-main.1054",
    pages = "18949--18954",
    abstract = "Argument relation classification (ARC) identifies supportive, contrasting and neutral relations between argumentative units. The current approaches rely on transformer architectures which have proven to be more effective than traditional methods based on hand-crafted linguistic features. In this paper, we introduce DISARM, which advances the state of the art with a training procedure combining multi-task and adversarial learning strategies. By jointly solving the ARC and discourse marker detection tasks and aligning their embedding spaces into a unified latent space, DISARM outperforms the accuracy of existing approaches.",
}

Installation

Disarm was tested on Python 3.10.12 using a Google Colab V100 system. It is recommended to set up the model in a virtual environment.

python -m venv [NAME_OF_ENV]
source [NAME_OF_ENV]/bin/activate

pip install -r requirements.txt

How to run the experiments

Unzip data.zip. This will create a new directory data/ containing the datasets StudentEssay ("student_essay"), Debatepedia ("debate") and M-ARG ("m-arg"). In the following sections, substitute [NAME_OF_DATASET] with one of the names in brackets.

Run adversarial training on ARC/DMD with grid search

python main.py --dataset [NAME_OF_DATASET] --adversarial --grid_search

Run adversarial training on a specific combination of loss weights

python main.py --dataset [NAME_OF_DATASET] --adversarial --discovery_weight [DISCOVERY_VALUE] --adv_weight [ADV_WEIGHT]

with [DISCOVERY_VALUE] and [ADV_WEIGHT] being scalar values in [0,1].

To run the tests on Multi-Task Learning (MTL), set the --adv_weight to 0.

Run RoBERTa+ finetuning

python main.py --dataset [NAME_OF_DATASET]

Visualize the embeddings

python main.py --dataset [NAME_OF_DATASET] --visualize [NAME_OF_DATASET_TO_PLOT]

where [NAME_OF_DATASET_TO_PLOT] must be either [NAME_OF_DATASET] or discovery In order to visualize the embeddings, you must have trained a model on the dataset you want to plot. After training, the best model is saved in a file called [NAME_OF_DATASET]_model.pt. The --visualize flag forces the program to load the model, calculate the embeddings of the dataset and plot them.