A Public Dataset Based on Mainstream Social Platforms

Weibo Dataset

V1: Includes data on 2,106 news items from the microblogging platform in the half of 2023. There are 1,000 fake news and 1,067 real news. The dataset consists of comment data on news spreads and contains user, comment, and multi-model information.

V2: Includes 11,329 number of news from the Chinese microblogging social media platform. There are 5,661 fake news items and 5,668 real news items. Comparable to version 1 (V1), version 2 (V2) expands the data magnitude on the basis of V1. Meanwhile, V2 provides news multi-modal data, including news posts, comment collections, images, videos and voice information. As a result, V2 provides a better simulation of the real environment of social networks, thus supporting downstream tasks.

V3: In progress. Please wait. The next version of the larger dataset is expected to be released in the latter half of 2025.

Datsets link: https://github.com/yzhouli/SocialNet/tree/master/Weibo.

TikTok Dataset

V1: Includes multimodal event propagation features. Moreover, the TikTok dataset can provide real data support for downstream tasks.

V2: In progress. Please wait. The first version of the TikTok dataset is expected to be released in 2025.

Twitter (X) Dataset

In progress. Please wait.

Please Cite

Source Paper: Yang Z, Pang Y, Li Q, et al. A model for early rumor detection base on topic-derived domain compensation and multi-user association[J]. Expert Systems with Applications, 2024: 123951. [Online]. Available: https://doi.org/10.1016/j.eswa.2024.123951.

@article{yang2024model,
  title={A model for early rumor detection base on topic-derived domain compensation and multi-user association},
  author={Yang, Zhou and Pang, Yucai and Li, Qian and Wei, Shihong and Wang, Rong and Xiao, Yunpeng},
  journal={Expert Systems with Applications},
  pages={123951},
  year={2024},
  publisher={Elsevier}
}