Using Pre-Trained Language Models in an End-to-End Pipeline for Antithesis Detection

This repository contains the official implementation of "Using Pre-Trained Language Models in an End-to-End Pipeline for Antithesis Detection" accepted in LREC-2024 with the newly proposed antithesis dataset (antithesis_dataset.csv) publicly available

Installation

Run command below to install the environment (using python3.9):

pip install -r requirements.txt

Fine-tuning

Run command below to fine-tune encoder language model, e.g., BERT for the antithesis detection task:

python antithesis_detection.py 

Results

On test set of the antithesis dataset:

Cite

@article{saadikuh2024Anti,
  title={Using Pre-trained Language Models in an End-to-End Pipeline for Antithesis Detection},
  author={Kuhn, Ramona, and Saadi, Khouloud and Mitrovic, Jelena and Granitzer, Michael},
  journal={LREC},
  year={2024}
}