FakeSV

Official repository for "FakeSV: A Multimodal Benchmark with Rich Social Context for Fake News Detection on Short Video Platforms", AAAI 2023.

• Dataset: The video ID (which can be used to infer the video URL) and corresponding annotations have been released. Also, we provide two data split used in the paper, i.e. event-based and temporal.
• Models: We reproduce some SOTA methods on fake news video detection to provide benchmark results for FakeSV. Codes for our proposed model SV-FEND and other methods are provided.

Environment

Anaconda 4.13.0, python 3.8.5, pytorch 1.10.1 and cuda 11.7. For other libs, please refer to the file requirements.txt.

Application for Data Use

Please sign this agreement and send the signed copy to [email protected].

Data Processing

video-subtitle-extractor

bert-base-chinese

VGG19

C3D

VGGish

Citation

@inproceedings{fakesv, 
title={FakeSV: A Multimodal Benchmark with Rich Social Context for Fake News Detection  on Short Video Platforms}, 
author={Qi, Peng and Bu, Yuyan and Cao, Juan and Ji, Wei and Shui, Ruihao and Xiao,  Junbin and Wang, Danding and Chua, Tat-Seng}, 
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence}, 
year={2023}, 
organization={AAAI} 
} 

Related Survey:

@article{fakesvsurvey,
  title={Online Misinformation Video Detection: A Survey},
  author={Yuyan Bu, Qiang Sheng, Juan Cao, Peng Qi, Danding Wang and Jintao Li},
  journal={arXiv preprint arXiv:2302.03242},
  year={2023}
}