Misinformation Detection in Russo-Ukrainian Conflict Tweets

This repository contains the code and resources for the ClimateSense submission to PROMID 2025 Subtask 3: "Misinformation Detection in Social Media Texts".

The objective of this task is to classify tweets related to the Russo-Ukrainian conflict as either misinformation (positive class) or non-misinformation (negative class). The dataset comprises manually annotated tweets, collected using the Twitter API during the first year of the Russia-Ukraine war.

Data

• The official PROMID 2025 Subtask 3 dataset can be found here: https://www.codabench.org/datasets/download/a923a05f-fcad-4a2f-815c-46c3233be44f. It includes:
  • misinfo_train.csv — labelled misinformation tweets
  • nonmisinfo_train.csv — labelled non-misinformation tweets
• Additional augmentation used in experiments: 5,022 Ukraine-related misinformation tweets from the Fact-checking Observatory (FCO). Note: this data is not available publicly due to licensing restrictions.

Reproducibility & Usage

To reproduce our experiments:

1. Clone this repository.
2. Download the official PROMID 2025 Subtask 3 dataset (see Data above) and place the files in the data/ directory as follows:

   data/
   ├── misinfo_train.csv
   └── nonmisinfo_train.csv
   

3. Open and run promid_task3.ipynb, which includes all steps for preprocessing, training, and evaluation for both baseline and augmented experiments.

License

See the LICENSE file in the repository for licensing information. For questions or collaboration, open an issue or contact the maintainers directly.

Acknowledgements

This work was supported by the European CHIST-ERA program within the ClimateSense project (Grant ID ANR-24-CHR4-0002, EPSRC EP/Z003504/1).