---
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
dataset_info:
  features:
  - name: text
    dtype: string
  - name: stance
    dtype: string
  - name: topic
    dtype: string
  splits:
  - name: train
    num_bytes: 3110594
    num_examples: 11500
  - name: test
    num_bytes: 818725
    num_examples: 2977
  download_size: 1724792
  dataset_size: 3929319
task_categories:
- text-classification
- zero-shot-classification
- feature-extraction
language:
- ar
tags:
- BERT
- MARBERT
- Scocial-Media
size_categories:
- 10K<n<100K
---
# ArabicStanceX
ArabicStanceX: A large-scale social media dataset for Arabic stance detection (14.5K tweets, 17 topics)
ArabicStanceX is a large-scale, richly annotated dataset for stance detection in Arabic social media text. It consists of 14,477 tweets covering 17 diverse topics across six domains: Sport, Education, Health, Religion, Economy, and Other.

Dataset Structure
The dataset is organized into two main splits:

train

test

Each of these splits contains six topics, one for each domain:

education

sport

health

religion

economy

other

Within each domain folder, there are topic-specific .json files.
Each .json file corresponds to a specific topic and contains a list of entries. Each entry includes:


text: The Arabic tweet content

stance: The annotated stance label (favor, against, or none)

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

This dataset is licensed under CC BY 4.0. You are free to share and adapt it with proper attribution.

# Citation
Alkhathlan, A., Alahmadi, F., Kateb, F., & Al-Khalifa, H. (2025). Constructing and evaluating ArabicStanceX: A social media dataset for Arabic stance detection. Frontiers in Artificial Intelligence, 8, Article 1615800