---
license: cc0-1.0
language:
- zh
size_categories:
- n<1K
task_categories:
- text-generation
- feature-extraction
tags:
- islam
- travel
- mosques
- halal
- raw-corpus
- rag
pretty_name: Salaamalykum Islamic Articles
configs:
- config_name: corpus
  data_files:
  - split: train
    path: data/corpus.parquet
- config_name: rag_chunks
  data_files:
  - split: train
    path: data/rag_chunks.parquet
---

# Salaamalykum Islamic Articles Corpus (Parquet)

This is a highly structured, machine-readable dataset comprising 34 original articles curated from [salaamalykum.com](https://salaamalykum.com). 
It has been natively formatted in **Parquet** to support efficient SQL querying in the Hugging Face Dataset Viewer and Data Studio.

## Strict Metadata Adherence
- **Dataset Hygiene**: Contains ONLY informational articles. No fake `ShareGPT` or `Alpaca` tags are used.
- **Data Provenance**: All articles are authored by Hasan09, tracking Muslim travel guides, mosque histories, and cross-sectarian observations.

## Views (Splits)
This repository exposes two subsets via Parquet formatting to prevent `TooBigContentError` and enable scalable RAG architectures:

### 1. `corpus` (Full Articles)
Designed for LLM pre-training and long-context analysis.
- `id`: Unique article identifier.
- `title`: Article title.
- `url`: Canonical URL.
- `word_count`: Length of the text for easy filtering in SQL.
- `tags`: Comma-separated categories.
- `preview`: Lightweight summary for UI viewers.
- `content`: The complete Markdown text.

### 2. `rag_chunks` (Retrieval-Augmented Generation)
Designed for Embedding models, Semantic Search, and Atlas visualizations.
- `chunk_id`: Unique chunk identifier.
- `text`: Specific paragraph.
- `chunk_index`: Position of the paragraph in the original text.

## Usage (SQL Console)
You can directly query this dataset via the Hugging Face SQL Console:
```sql
SELECT title, url FROM "corpus" WHERE word_count > 1000
```