Arabic Fake News Articles for Machine Learning Detection Models

A collection of Arabic news articles annotated for fake news detection research. The dataset was prepared by Abd, Dhafar for evaluating machine learning-based misinformation detection systems and is hosted on Harvard Dataverse. It has been cleaned and preprocessed to facilitate reproducible research in Arabic natural language processing.

Use Cases

• Train binary classifiers for fake news detection based on labeled article text.
• Benchmark the performance of new Arabic NLP models on misinformation detection tasks.
• Evaluate the adversarial robustness of text classification models using the prepared samples.
• Apply transfer learning from models trained on this dataset to related Arabic text tasks.

Strengths

• Articles are labeled for real and fake news, providing a clear target for supervised learning.
• Data has been cleaned and preprocessed, which may reduce initial processing overhead for researchers.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale model training.

Provenance

Source

Harvard Dataverse

Collection Method

Collected and annotated news articles, as described.

Freshness

Last updated 2026-07-10 14:22:06; freshness should be verified.