Datasets:

shimaa22
/

arabic_students_comments

language:
  - ar
license: mit
task_categories:
  - text-classification
task_ids:
  - sentiment-analysis
pretty_name: Egyptian Arabic Text Classification Dataset
size_categories:
  - small

📊 Egyptian Arabic Text Classification Dataset

🧾 Dataset Description

This dataset contains 1,353 Egyptian Arabic (colloquial) text samples labeled for text classification tasks.
Each sample consists of a short sentence describing institutional or educational features, annotated with a categorical label representing the strength or presence of the feature.

The dataset is suitable for Natural Language Processing (NLP) tasks such as:

Text Classification
Sentiment / Strength Analysis
Feature Evaluation Models

🔹 Labels are assigned based on students' opinions:

Positive opinions → strong

Negative opinions → weak

⚠️ Note: All text is in Egyptian Arabic (عامية مصرية), which may require special preprocessing or tokenization.

📁 Data Structure

The dataset consists of two columns:

Column	Description
`text`	Egyptian Arabic textual description
`label`	Classification label (`strong` or `weak`)

✏️ Example Samples

text	label
الجو في المعهد مريح وهادي	strong
البرامج التعليمية مش متنوعة	weak
الموقع الإلكتروني موجود وبيشتغل كويس	strong

🏷️ Labels

strong: Positive or strong presence of the described feature.
weak: Negative or weak presence of the described feature.

⚠️ Future versions of the dataset may include additional labels.

🎯 Intended Use

Training and evaluating Arabic NLP classification models
Academic research in educational quality assessment
Experiments with transformers and deep learning models
Benchmarking Egyptian Arabic text understanding

⚠️ Limitations

The dataset is relatively small (1,353 samples) and may require augmentation for deep learning models.
Labels are domain-specific and reflect a particular evaluation context.
Text is in Egyptian Colloquial Arabic, not Modern Standard Arabic.
Not intended for high-stakes or real-world decision-making.

⚖️ Ethical Considerations

The dataset does not contain personal or sensitive information.
All texts are generic and institution-related.
The dataset is provided for research and educational purposes only.

📜 License

MIT License

📌 Citation

Mahmoud, A. A. (2026). Egyptian Arabic Text Classification Dataset [Data set]. Retrieved from Hugging Face Hub.