Arabic Machine-Generated Social Media Posts Dataset

This dataset contains machine-generated Arabic social media posts using a text polishing approach across multiple Large Language Model (LLMs). It was created as part of the research paper: "Arabic machine-generated text detection: Stylometric analysis and cross-model evaluation" (https://www.sciencedirect.com/science/article/abs/pii/S0957417425042599).

📋 Dataset Overview

The dataset addresses the need for comprehensive Arabic machine-generated text resources in informal/social media contexts, enabling research in detection systems, stylometric analysis, and cross-model generalization studies in casual Arabic writing.

Generation Method

by_polishing - Text refinement approach where models polish existing human social media posts while preserving dialectal expressions, diacritical marks, and the informal nature of social media writing.

Data Fields

Each sample contains:

• original_post: The original human-written Arabic social media post
• {model}_generated_post: Machine-generated polished version from each model
  • allam_generated_post
  • jais_generated_post
  • llama_generated_post
  • openai_generated_post

🤖 Models Used

📊 Dataset Statistics

Dataset Overview

Content Statistics

Data Collection & Preprocessing

The dataset was constructed from two prominent Arabic review collections: BRAD (Book Reviews in Arabic Dataset) collected from Goodreads.com and HARD (Hotel Arabic Reviews Dataset) collected from Booking.com. Both datasets primarily contain Modern Standard Arabic text with loose language and informal tone. The selection focused on obtaining longer-form reviews suitable for meaningful linguistic analysis. From BRAD, 3,000 reviews were selected containing between 724-1,500 words per review, while from HARD, 500 reviews ranging from 150-614 words were extracted. Preprocessing steps included removal of special characters and non-printable text, normalization of Arabic text through tatweel removal, and standardization of repeated punctuation marks (limiting repetitions to a maximum of 3). After generation, we filter samples by removing invalid generated posts, setting a minimum threshold of 50 words per post, and dropping duplicated samples (if there are any). The resulting final dataset after this filtration contains 3,318 samples. What is currently available in this dataset repository represents the final processed outcome of this entire pipeline. For access to the original papers, metadata, and links, please visit The paper repo (requires cloning with Git LFS enabled due to large file sizes). For further details, please refer to the full paper.

Generated Text Length by Model

🔍 Highlighted Findings from the Research Paper

These are only spotlights, please refer to the paper for more details.

Social Media Context Insights

• Notable length reduction across all models (26-72% of human length)
• ALLaM shows closest alignment to human writing patterns in informal contexts
• Higher semantic preservation (BERTScores 81.7-95.1%) despite length differences

Detection Performance in Social Media

• Challenging detection
• Cross-model generalization degradation
• Model-specific detectability variations: Llama (95.4% F1) > Jais (87.0%) > OpenAI (81.5%) > ALLaM (66.5%)

🚀 Getting Started

Loading the Dataset

from datasets import load_dataset

# Load the complete dataset
dataset = load_dataset("KFUPM-JRCAI/arabic-generated-social-media-posts")

# Example: Get a sample
sample = dataset[0]
print("Original:", sample["original_post"])
print("ALLaM:", sample["allam_generated_post"])
print("Jais:", sample["jais_generated_post"])

Practical Applications

• Social media content verification
• Informal Arabic AI detection tools
• Cross-domain detection system evaluation
• Casual writing style analysis

📚 Citation

If you use this dataset in your research, please cite:

The Expert Systems with Applications Journal paper:

@article{al2025arabic,
  title={Arabic Machine-Generated Text Detection: Stylometric Analysis and Cross-Model Evaluation},
  author={Al-Shaibani, Maged S and Ahmed, Moataz},
  journal={Expert Systems with Applications},
  pages={130644},
  year={2025},
  publisher={Elsevier}
}

The preprint (arxiv):

@article{al2025arabic,
  title={The Arabic AI Fingerprint: Stylometric Analysis and Detection of Large Language Models Text},
  author={Al-Shaibani, Maged S and Ahmed, Moataz},
  journal={arXiv preprint arXiv:2505.23276},
  year={2025}
}

🏢 Institutional Support

This work was supported by:

• SDAIA-KFUPM Joint Research Center for Artificial Intelligence

⚖️ Ethical Considerations

This dataset is intended for research purposes to:

• Improve detection of machine-generated content in informal contexts
• Enhance social media integrity tools
• Advance Arabic NLP research in casual writing
• Understand LLM behaviors in informal Arabic contexts

Please use responsibly and in accordance with your institution's research ethics guidelines.

🔗 Related Datasets

• Academic Abstracts: KFUPM-JRCAI/arabic-generated-abstracts
• Source Datasets: BRAD (Book Reviews), HARD (Hotel Reviews)

🔄 Updates and Versions

• v1.0: Initial release with social media posts using polishing generation method
• Check the dataset card for the latest updates

---