ijdutse/hausa-misinformation-dataset

A curated dataset of annotated fake & genuine news in Hausa. Resources for NLP, misinformation detection, and media analysis in low-resource languages. # Hausa Misinformation Dataset A curated and annotated dataset of fake and genuine online content in the Hausa language. This resource is intended to support research in Natural Language Processing (NLP), misinformation detection, and media studies for low-resource languages. ## 📥 Download Data The latest version of the dataset can be found in the `/data/processed/` directory. - **Primary File**: `hausa_misinformation_dataset_v1.0.csv` - **Format**: CSV (Comma-Separated Values) - **Encoding**: UTF-8 **Direct Download Link**: Click here to download v1.0 *(Alternatively, you can use Git LFS if the dataset is very large)* ## 🗂️ Dataset Overview | Column Name | Description | Data Type | | :--- | :--- | :--- | | `id` | Unique identifier for each text sample | Integer | | `text` | The full text of the news article/social media post | String | | `label` | Annotation label (`fake` or `genuine`) | Categorical | | `source` | Type of source (e.g., `facebook`, `online_news`, `twitter`) | Categorical | | `date_collected` | Date when the sample was collected | Date | | `annotator_notes` | Optional notes from the annotators | String | **Statistics:** - Total samples: ~1000 - Fake samples: ~500 - Genuine samples: ~500 - Average text length: 250 words ## 🚀 Quick Start 1. **Clone the repository** ```bash git clone github.com cd hausa-misinformation-dataset ``` 2. **Load the data in Python (Pandas)** ```python import pandas as pd df = pd.read_csv('data/processed/hausa_misinformation_dataset_v1.0.csv') print(df['label'].value_counts()) ``` 3. **Explore the data** Check out the Jupyter notebooks in the `/notebooks/` directory for examples of data analysis and a baseline model. ## 📚 Documentation For detailed information, please refer to our documentation: - **Data Collection Methodology**: How the data was sourced and gathered. - **Annotation Guidelines**: The criteria and process used by annotators to label content. - …