political_classifier's Introduction

Political Classifier

A fine-tuned RoBERTa model for classifying text as either political (1) or non-political (0).

MODEL USAGE:

Download the pretrained model, unzip it, and place the folder in the saved_models/ directory
Install requirements using pip install -r requirements.txt (I would recommend creating a virtual environment before doing so)
Your input data needs to be a csv with a text column and the model can then generate predictions for it-- a sample data file is included in the repository as sample_data.csv
To generate predictions (if your dataset is named sample_data.csv, otherwise change accordingly), run the following command inside the main directory: python test_model.py -m saved_models/pol_clf_checkpoint -d sample_data.csv
You should now have a predictions.csv file with a classifier output column named prediction
If you need the checkpoint files please reach out by email to anshumanc[at]usf[dot]edu.

TRAINING DATA:

Available here

CITATION:

If you use our model/data, please cite the following papers:

Plain-text:

Askari, H., Chhabra, A., von Hohenberg, B. C., Heseltine, M., & Wojcieszak, M. (2024). Incentivizing News Consumption on Social Media Platforms Using Large Language Models and Realistic Bot Accounts. arXiv preprint arXiv:2403.13362.

Bib:

title={Incentivizing News Consumption on Social Media Platforms Using Large Language Models and Realistic Bot Accounts},
author={Askari, Hadi and Chhabra, Anshuman and von Hohenberg, Bernhard Clemm and Heseltine, Michael and Wojcieszak, Magdalena},
journal={arXiv preprint arXiv:2403.13362},
year={2024}}

Recommend Projects