license: cc-by-4.0
📰 multiCred: A Dataset of News Organization Credibility on Twitter
multiCred is a research dataset collected and used in the paper
“Multilevel User Credibility Assessment in Social Networks.”
It contains Twitter (X) accounts belonging to news organizations and their associated tweet and comment IDs.
The dataset is designed to support studies in media credibility, information reliability, and news dissemination patterns on social media.
This dataset follows the Twitter Developer Agreement and Policy.
It includes only Tweet IDs and User IDs — no tweet text, usernames, or personal data are shared.
🧩 Dataset Summary
| Aspect | Description |
|---|---|
| Paper | Multilevel User Credibility Assessment in Social Networks |
| Entities | 1,233 Twitter accounts belonging to news organizations |
| Fields per entry | user, tweets, comments, label_available |
| Data Type | Tweet and User IDs only |
| Labels | Some accounts have credibility labels from NewsGuard (not publicly shared) |
| Language | Various (depending on the organization) |
| License | CC BY-NC 4.0 |
| Use Case | Research on news credibility, media bias, and misinformation |
💾 Data Structure
Each entry in multiCred.json follows this structure:
{
"user": "1234567890",
"tweets": ["111111111111111111", "222222222222222222"],
"comments": ["333333333333333333", "444444444444444444"],
"label_available": true
}
📰 Dataset Composition
- Accounts: 1,233 verified or recognized news organizations
- Tweets: IDs of posts authored by these organizations
- Comments: IDs of replies directed at those accounts
- Labels: Some organizations are labeled according to NewsGuard’s credibility ratings (labels not publicly released)
⚖️ Ethical and Legal Considerations
- All accounts represent institutional, public news outlets, not private individuals.
- Only Tweet IDs and User IDs are shared — no tweet text, usernames, bios, or images.
- The dataset fully complies with:
- Twitter Developer Policy
- NewsGuard Terms of Use
🔒 Access to Credibility Labels
- A portion of this dataset includes credibility labels derived from NewsGuard.
- Due to copyright restrictions, these labels cannot be publicly shared.
- Researchers interested in accessing the labeled subset may contact for non-commercial, academic research purposes.
Contact: [mohammad.moradi9775@gmail.com]
Please include your affiliation and intended use.
🧰 How to Rehydrate Tweets
You can rehydrate Profile IDs, Tweet IDs and IDs of replies to obtain profile, tweet text and their metadata using tools like twarc2 and snscrape.
🧠 Intended Uses
- Studying news credibility and trustworthiness on social media
- Modeling information diffusion from credible vs. non-credible sources
- Developing and evaluating automatic credibility assessment systems
🚫 Limitations
- The dataset does not include tweet text or NewsGuard ratings
- Some accounts may be inactive or removed since data collection
- NewsGuard data remains proprietary and cannot be redistributed
- Tweet IDs may no longer resolve if the tweet or account has been deleted
- Labels are only available for a subset of accounts and must be requested separately
🪪 License
- License: CC BY-NC 4.0
- You may share and adapt the dataset for non-commercial research with proper attribution, and in full compliance with Twitter and NewsGuard policies
🧾 Citation
Please cite both the dataset and the original paper:
@dataset{multicred_2025,
author = {Mohammad Moradi},
title = {multiCred: A Dataset of News Organization Credibility on Twitter},
year = {2025},
publisher = {Hugging Face Datasets},
url = {https://huggingface.co/datasets/mamad97/MultiCred-Dataset}
}
@article{DBLP:journals/corr/abs-2309-13305,
author = {Mohammad Moradi and
Mostafa Haghir Chehreghani},
title = {Multilevel User Credibility Assessment in Social Networks},
journal = {CoRR},
volume = {abs/2309.13305},
year = {2025},
url = {https://doi.org/10.48550/arXiv.2309.13305},
doi = {10.48550/ARXIV.2309.13305},
}