qurancn/Islamic-Culture

☪ Islamic-Culture: Chinese Islamic Knowledge Base & RAG Corpus

 

[!TIP] Human Readers: Looking for the full text with all images perfectly rendered? Navigate to the [Files and versions -> content](https://huggingface.co/datasets/qurancn/Islamic-Culture/tree/main/content) folder to browse all Markdown articles natively!

Dataset Description

Islamic-Culture is a curated Retrieval-Augmented Generation (RAG) corpus containing 260 native Chinese articles covering Islamic culture, heritage, architecture, halal food, and Muslim community life across China, Southeast Asia, the Middle East, and beyond.

Key Properties

Embeddable Dataset Viewer

You can embed the dataset viewer in any webpage:

<iframe
  src="https://huggingface.co/datasets/qurancn/Islamic-Culture/embed/viewer/default/train"
  frameborder="0"
  width="100%"
  height="560px"
></iframe>

Direct viewer link: 🔍 Browse Dataset

Filter by topic: Islamic Architecture | Halal Food | Travel Guide

Topic Distribution

Data Schema

SQL Console Queries

You can run SQL directly in Data Studio:

-- Count articles by topic
SELECT topic_category, COUNT(*) as count
FROM train
GROUP BY topic_category
ORDER BY count DESC;

-- Find longest articles
SELECT title, char_count, source_url
FROM train
ORDER BY char_count DESC
LIMIT 10;

-- Search for mosque-related articles
SELECT title, text_preview, source_url
FROM train
WHERE title LIKE '%清真寺%';

Data Samples

Example: 1550年的伊斯俩目世界之旅（第一篇）

Example: 1550年的伊斯俩目世界之旅（第三篇）——南亚

Example: 1550年的伊斯俩目世界之旅（第二篇）——蒙古帝国的遗产（上篇）

Data Collection & Processing Pipeline

Source

All articles originate from salaamalykum.com, a community-driven platform documenting Chinese Muslim life and Islamic culture.

Extraction Method

Articles were extracted directly from the source platform's MySQL database (aws_article table, topic_id=745), ensuring complete content retrieval without web-scraping artifacts or pagination bugs.

Processing Steps

1. 1.SQL Parsing: Custom state-machine parser reads MySQL dump, extracts tuples from INSERT INTO statements
2. 2.Topic Mapping: aws_topic_relation table maps articles to topic ID 745 (伊斯兰文化)
3. 3.BBCode Conversion: [img]...[/img] → ![image](...), [b]...[/b] → **...**
4. 4.HTML Cleanup: html.unescape() + <br> normalization
5. 5.Integrity Hashing: SHA-256 computed on cleaned body text
6. 6.Category Classification: Rule-based keyword classifier assigns one of 6 topic categories
7. 7.Parquet Serialization: pyarrow with row_group_size=50, write_page_index=True for optimal Viewer performance

Known Limitations

• —No machine translation: All text is native Chinese — this is intentional and a core feature
• —Image links: Articles reference external image URLs (archive.org hosted); images are not embedded in the dataset
• —Author attribution: Individual author names are not preserved (WeCenter platform limitation); all attributed to "Salaamalykum User"
• —Temporal coverage: Articles span approximately 2024–2026

Usage

Python (Hugging Face Datasets)

from datasets import load_dataset

ds = load_dataset("qurancn/Islamic-Culture")
print(ds)
# DatasetDict({
#     train: Dataset({features: [...], num_rows: 208})
#     validation: Dataset({features: [...], num_rows: 26})
#     test: Dataset({features: [...], num_rows: 26})
# })

# Filter by topic
mosques = ds["train"].filter(lambda x: x["topic_category"] == "Islamic Architecture")

LangChain RAG

from langchain.vectorstores import FAISS
from langchain.embeddings import HuggingFaceEmbeddings
from datasets import load_dataset

ds = load_dataset("qurancn/Islamic-Culture", split="train")
texts = [f"{row['title']}\n{row['text']}" for row in ds]

embeddings = HuggingFaceEmbeddings(model_name="shibing624/text2vec-base-chinese")
vectorstore = FAISS.from_texts(texts, embeddings)

Version History

Citation

@dataset{islamic_culture_2026,
  title={Islamic-Culture: A Curated Chinese RAG Dataset of 260 Articles on Islamic Culture},
  author={Salaamalykum Project},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/qurancn/Islamic-Culture}
}

License

This dataset is released under the MIT License.

Contact

• —Email: bropeace@protonmail.com
• —GitHub: https://github.com/salaamalykum
• —Website: https://salaamalykum.com