Moral-Role-Prediction

This repository contains dataset and codes for predicting subframes in text. The approach is describe in the following paper.

Identifying Morality Frames in Political Tweets using Relational Learning
Shamik Roy, Maria Leonor Pacheco and Dan Goldwasser
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP 2021)

Dataset

The annotated tweet dataset can be found in the 'annotated_dataset' folder. The folder contains separate json files for each of the Moral Foundations. Upon parsing the json files a dictionary of the following format can be found.

{
tweet_id:
  {
  'annotations': ...,
  'author-label': ..., 
  'dop': ..., 
  'issue': ..., 
  'text': ...
  }
}

tweet_id : Actual tweet id
annotations : Character indexed annotation by each of the annotators
author-label : Political affiliation of the author of the tweet (Republican or Democrat)
dop : Date of Publication
issue : Topic of the tweet
text : tweet text

The 'dataset_description.txt' file contains more information about the data files. Note that, we cannot release the tweet texts because of Twitter Privacy Policy. The tweet text can be parsed using the tweet_ids provided in the json files. If you need those immediately or require any help in parsing feel free to send an e-mail to roy98@purdue.edu .

Code

Coming soon. Please send an e-mail to roy98@purdue.edu for emergency cases.

Citation

If you find the dataset and approach helpful in your work, please cite the paper.

@inproceedings{roy2021identifying,
  title={Identifying Morality Frames in Political Tweets using Relational Learning},
  author={Roy, Shamik and Pacheco, Mar{\'\i}a Leonor and Goldwasser, Dan},
  booktitle={Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing},
  pages={9939--9958},
  year={2021}
}