EANN: Event Adversarial Neural Networks for Multi-Modal Fake News Detection
Yaqing Wang,
Fenglong Ma,
Zhiwei Jin, Ye Yuan,
Guangxu Xun,
Kishlay Jha,
Lu Su,
Jing Gao
SUNY Buffalo. KDD, 2018.
The data folder contains the partial dataset. The train_id, validate_id and test_id are the event id dictionaries.
The src folder contains the data preprocessing file: process_data_weibo.py, and the model files: EANN.py and EANN_text.py. EANN.py is for text and image multimodal features and EANN_text.py is only using textual featues.
Note: This code is written in Python2.
python EANN.py or python EANN_text.py
One of the unique challenges for fake news detection on social media is how to identify fake news on newly emerged events. The EANN is desgined to extract shared features among all events to effectively improve the performance of fake news detection on never-seen events.
Comparision between reduced model (w/o adversarial) and EANN(w adversarial)
The feature representations learned by the proposed model EANN (right) are more discriminable than fake news detection (w/o adv).
If you use this code for your research, please cite our paper:
@inproceedings{wang2018eann,
title={EANN: Event Adversarial Neural Networks for Multi-Modal Fake News Detection},
author={Wang, Yaqing and Ma, Fenglong and Jin, Zhiwei and Yuan, Ye and Xun, Guangxu and Jha, Kishlay and Su, Lu and Gao, Jing},
booktitle={Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery \& Data Mining},
pages={849--857},
year={2018},
organization={ACM}
}