Tarek Mahmoud1, Zhuohan Xie1, Dimitar Dimitrov2, Nikolaos Nikolaidis3, Purificação Silvano4, Roman Yangarber5, Shivam Sharma6, Elisa Sartori7, Nicolas Stefanovitch8, Giovanni Da San Martino7, Jakub Piskorski9, Preslav Nakov1
1MBZUAI (Mohamed bin Zayed University of AI), Abu Dhabi, UAE
2Sofia University “St. Kliment Ohridski”, Sofia, Bulgaria
3Athens University of Economics and Business, Athens, Greece
4University of Porto, Porto, Portugal
5University of Helsinki, Helsinki, Finland
6Indian Institute of Technology Delhi, New Delhi, India
7University of Padova, Padova, Italy
8European Commission Joint Research Centre, Ispra, Italy
9Institute of Computer Science, Polish Academy of Sciences, Warsaw, Poland
The dataset used in this paper was also part of the SemEval 2025 Task 10 shared task. To download the dataset, please register using the provided link.
The following files were used for our experiments:
target_4_December_release.zip
cleaned_dev_10_january_2025.zip
For our experiments, we create the train/dev sets by splitting the data in target_4_December_release.zip
based on unique article IDs (80/20 split, seed = 42) into train and dev sets. The test set in the paper is exactly the contents of cleaned_dev_10_january_2025.zip
.
When you register and download the dataset, you will receive data for 3 subtasks (1, 2, and 3). This paper uses only the Subtask 1 data; you may ignore the data for other subtasks.
Baselines and scoring tools are also provided on the registration page to help you understand the dataset structure and format.
Extended dataset: The link also provides additional data beyond what was used in this paper. These files are part of the extended dataset:
training_data_RU_final_19_January_2025_release.zip
testdata_ST12.zip
If you use our work, please cite:
@misc{mahmoud2025entityframingroleportrayal,
title={Entity Framing and Role Portrayal in the News},
author={Tarek Mahmoud and Zhuohan Xie and Dimitar Dimitrov and Nikolaos Nikolaidis and Purificação Silvano and Roman Yangarber and Shivam Sharma and Elisa Sartori and Nicolas Stefanovitch and Giovanni Da San Martino and Jakub Piskorski and Preslav Nakov},
year={2025},
eprint={2502.14718},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2502.14718},
}