Datasets-for-Information-Diffusion-Tasks

This is a collection of datasets for information diffusion tasks for A Survey of Datasets for Information Diffusion Tasks. The paper can be downloaded here.

📌Table of Content

• Datasets for Information Diffusion Tasks
  • 🔥Survey
  • 📊Datasets
  • Citation
  • License

Survey

We survey a systematic taxonomy of information diffusion tasks and datasets based on the "5W Model" framework.

• We categorize the information diffusion tasks into ten subtasks with definitions, datasets analysis and representative methodologies, from three main tasks of information diffusion prediction, social bot detection, and misinformation detection.

• We collect the publicly available dataset repository of information diffusion tasks with the available links and compare them based on six attributes affiliated to users and content: user information, social network, bot label, propagation content, propagation network, and veracity label.

• We discuss the limitations and future directions of current datasets and research topics to advance the future development of information diffusion.

Figure 1: Categorization of information diffusion tasks under the "5W Model" framework includes information diffusion prediction, social bot detection and misinformation detection, along with their respective subtasks and interrelationships.

Six attributes

• Attributes for user:

  • User information: Each user $u$ has their user information, including a basic user profile with a combination of semantic information (user description) and property information (numerical and categorical data), and optionally user historical information (posts and user engagements).
  • Bot label: $y^U$ denotes a social bot user (1) or a human user (0).
  • Social network: Among the user nodes set $V^U$, static relationships are denoted as the edges set $E^a$, such as follower/followee relationships, friendships, and co-authorships, which constitute the social network $G^a$.

• Attributes for content:

  • Propagation content: Each content $p$ published by users through media, such as posts, news, papers and comments, has its propagation content including texts/images/videos, tags, URLs, etc.
  • Veracity label: $y^P$ denotes false information (1) or true information (0). In rumor detection, this label typically indicates rumor ($R$) or non-rumor ($N$), and can also specify non-rumor ($N$), true rumor ($T$), false rumor ($F$), or unverified rumor ($U$).
  • Propagation network: Within a propagation content, dynamic relationships revealing propagation paths are generated by user engagement behaviors among content or user nodes $V={V^U, V^P}$, denoted as the edges set $E^b$, which constitute the propagation network $G^b$.

NB: Propagation network edges $E^b$ can be further divided into the following three types based on whether nodes represent users or contents in different tasks.

1. For user-to-user interactions in information diffusion prediction, the edges $E^{b1}$ represent the interaction relationships (click/like/repost/comment/citation) between two users $u \in V^U$ who publish the propagation content.

2. For content-to-content interactions in rumor detection, the edges $E^{b2}$ represent the interaction relationships (repost/comment/citation) between a post $p \in V^P$ and its repost/comment or between two reposts/comments.

3. For content-to-user interactions in fake news detection, the edges $E^{b3}$ represent the engagement relationships (click/like/repost/comment) between a news item $p \in V^P$ and its engagement elements such as users or between two engagement elements.

Comparison of datasets

In terms of the classification of tasks and the six attributes of data, we collect sixty-five publicly available datasets used in current works for each subtask, as shown in Table 2. Additionally, the available URLs for these datasets are provided in Table 3.

Datasets

Format:
- **Dataset name**
  - Source platforms of the dataset.
  - Click to download the dataset.
  - Paper: Source paper of the dataset.
  - Time horizon (if any)

Information Diffusion Prediction

Macroscopic Diffusion Prediction

Microscopic Diffusion Prediction

Social Bot Detection

Misinformation Detection

Citation

Our survey paper is "A Survey of Datasets for Information Diffusion Tasks" at arXiv, which includes more detailed discussions and we will continuously update it as well. The latest version was updated on July 6, 2024.

If you find our repo is helpful, please consider cite it.

@misc{guo2024survey,
      title={A Survey of Datasets for Information Diffusion Tasks}, 
      author={Fuxia Guo and Xiaowen Wang and Yanwei Xie and Zehao Wang and Jingqiu Li and Lanjun Wang},
      year={2024},
      eprint={2407.05161},
      archivePrefix={arXiv},
      primaryClass={cs.SI},
      url={https://arxiv.org/abs/2407.05161}, 
}

License

Datasets for Information Diffusion Tasks is released under the Apache 2.0 license.