AuthenticIlm/Shamela4_Full_DB

Shamela 4 — Full Islamic Library Corpus

A complete extraction of al-Maktaba al-Shamela (الشاملة) v4, containing 8,589 books across 40 categories of classical Islamic sciences. Extracted from the original Lucene + Sqlite Shamela DB on 2026-04-26 with ~7.6 million pages and ~19 GB of Arabic text.

Dataset Structure

stage0_raw/
├── _meta/                          # Cross-cutting metadata (Parquet + JSONL)
│   ├── extraction_manifest.json    # Global extraction record
│   ├── authors.parquet             # 3,187 authors
│   ├── book_metadata.parquet       # All 8,589 book records
│   ├── categories.parquet          # 41 categories
│   ├── quran_verses.parquet        # 6,236 verses (entire Quran)
│   ├── narrators.parquet           # 18,989 hadith narrators
│   ├── root_dictionary.parquet     # ~1.95M Arabic root entries
│   ├── hadith_xrefs.parquet        # ~37K cross-references
│   ├── tafsir_xrefs.parquet        # ~84K exegesis links
│   ├── page_isnads.parquet         # ~35K transmission chains
│   └── *.jsonl                     # Text-preview equivalents
│
├── 01__العقيدة/                    # Category directories (numbered 01–40)
│   └── 1__الفواكه-العذاب-في-الرد/  # Per-book directories
│       ├── manifest.json           # Integrity checksums + metadata
│       ├── book_metadata.json      # Bibliographic record
│       ├── toc.jsonl               # Table of contents (JSONL)
│       └── pages.jsonl             # Full page text (JSONL)
│
├── ...
│
└── 40__علوم-أخرى/

Per-Book Files

File	Format	Description
manifest.json	JSON	Extraction metadata, SHA-256 checksums, page/toc counts
book_metadata.json	JSON	Title, author(s), death dates, category, edition info
toc.jsonl	JSONL	Hierarchical table of contents with page anchors
pages.jsonl	JSONL	Full page content: body text, footnotes, page numbers

Pages JSONL Schema

{
  "page_id": 1,                    // Sequential page ID
  "book_id": 1,                    // FK to book
  "shamela_page_id": 1,           // Original Shamela ID
  "part": null,                    // Volume/part (for multi-volume books)
  "page_num": 3,                   // Printed page number
  "sequence_num": 1,              // Reading order
  "body": "Arabic text...",       // Main text with minimal HTML:
                                  // <span data-type='title' id=toc-N>Heading</span>
  "footnotes": null,              // Footnote text (¬١, ¬٢ markers)
  "hints": null,                  // Additional annotations
  "services_raw": null            // Source service data
}

TOC JSONL Schema

{
  "title_id": 1,
  "book_id": 1,
  "page_id": 5,                   // Page where heading appears
  "parent_id": null,              // Hierarchical parent (null = root)
  "shamela_title_id": 2,
  "title_text": "مقدمة المحقق"    // Heading text
}

Book Metadata Schema

{
  "book_id": 1,
  "shamela_id": 1,
  "title_ar": "الفواكه العذاب في الرد على من لم يحكم السنة والكتاب",
  "book_type": 1,
  "book_type_label": "كتاب",
  "category_id": 1,
  "category_name_ar": "العقيدة",
  "main_author_id": 513,
  "main_author_name_ar": "حمد بن ناصر آل معمر",
  "main_author_death_hijri": 1225,
  "authors": [
    { "author_id": 513, "role": "author", "name_ar": "...", "death_hijri": 1225 }
  ],
  "printed": true,
  "is_hidden": false,
  "betaka_text": "Full bibliographic description...",
  "version_major": 5,
  "version_minor": 0,
  "volume_count_observed": 0,
  "has_multi_part": false
}

Statistics

Metric	Count
Books	8,589
Pages	7,611,186
TOC entries	4,060,485
Authors	3,187
Categories	41
Isnad (chains)	35,526
Narrators	18,989
Quran verses	6,236
Tafsir cross-refs	84,430
Hadith cross-refs	37,076
Root dictionary	1,952,803
Total size	~19 GB
Meta directory	~177 MB
Rejected books	0

Largest Categories

#	Category	Books
06	كتب السنة (Hadith Collections)	1,241
01	العقيدة (Creed)	807
23	الرقائق والآداب والأذكار (Spirituality)	625
26	التراجم والطبقات (Biographies)	579
19	مسائل فقهية (Fiqh Issues)	429
32	الأدب (Literature)	406
39	كتب عامة (General)	357
10	علوم الحديث (Hadith Sciences)	320

Coverage

All major Islamic disciplines:

• Quranic: Tafsir, Quran Sciences, Tajwid & Qira'at
• Hadith: The six canonical collections (Kutub al-Sittah), Musnad Ahmad, Muwatta Malik, plus commentaries, sciences (ilal, rijāl, ilal), takhrij
• Fiqh: All four Sunni madhahib (Hanafi, Maliki, Shafi'i, Hanbali), general fiqh, usul al-fiqh, fiqh maxims, fatāwā
• Aqeedah: Classical creedal works, refutations, theological debates
• Arabic Language: Grammar (nahw), morphology (sarf), rhetoric (balāgha), lexicons, poetry, prosody
• History & Biography: Sirah, tarikh, tabaqat, genealogy, geography
• Other: Logic, medicine, general works

Data Notes

• Text is raw — minimal HTML markup preserved (<span data-type='title'>, <hr>, <table>)
• Footnote markers — (¬N) notation linking body to footnotes
• Quranic text — bracketed with ﴿...﴾ markers
• Honorifics — ﷺ (SAW), ﵁, ﷿, etc. are single Unicode codepoints
• Multi-volume works — linked via parent_id / group_id in metadata
• Author death dates — Hijri years; negative values = CE dates
• Hidden books — is_hidden: true indicates copyright/access restrictions
• \r line endings — Shamela uses \r (carriage return) instead of \n
• Base64 images — some pages contain inline <img> with base64 data

Extraction

• Source: Production PostgreSQL 16 (dockerized)
• Method: psycopg server-side cursor stream, constant memory
• Date: 2026-04-26
• Extractor: shamela-extractor v0.3.0
• Snapshot: shamela4_full_extraction_pg_2026_04_26
• No books rejected — all 8,589 books extracted successfully

Usage Examples

Load with Hugging Face Datasets

from datasets import load_dataset

# Load a single book by filtering
ds = load_dataset("AuthenticIlm/Shamela4_Full_DB", split="full")
book_1 = ds.filter(lambda x: x["book_id"] == 1)

# Or stream individual JSONL files directly
import json
with open("01__العقيدة/1__الفواكه-العذاب/pages.jsonl") as f:
    for line in f:
        page = json.loads(line)
        print(page["body"][:200])

Load metadata with Polars

import polars as pl

df = pl.read_parquet("_meta/book_metadata.parquet")
df.filter(pl.col("category_id") == 3)  # All tafsir books

Query roots from the root dictionary

df = pl.read_parquet("_meta/root_dictionary.parquet")
df.filter(pl.col("token").str.contains("صلو"))

Intended Uses

• Classical Islamic text research — full-text search, citation extraction, cross-book linking
• Arabic NLP — language modeling on classical Arabic (pre-modern, diacritized, diverse genres)
• Hadith studies — chain-of-transmission analysis, narrator networks, takhrij
• Intertextuality analysis — quotation detection, author attribution, topic propagation
• Lexicography — root extraction, semantic field analysis from medieval dictionaries
• LLM training/evaluation — high-quality classical Arabic data for fine-tuning or evaluation

Limitations

• This is a full DB dump, including the encoded Narrator tables, etc.

License

This dataset is derived from al-Maktaba al-Shamela (الشاملة), a free digital library of classical Islamic texts. The original texts are in the public domain. The compilation and digitization rights belong to their respective owners.

Terms of use: Research and personal use. Respect the intellectual property rights of muhaqqiqīn (editors) and publishers for critical editions.

Citation

@misc{shamela4_full_2026,
  title = {Shamela 4 — Full Islamic Library Corpus},
  author = {{AuthenticIlm}},
  year = {2026},
  note = {Extracted from al-Maktaba al-Shamela v4. 8,589 books, 7.6M pages.},
  url = {https://huggingface.co/datasets/AuthenticIlm/Shamela4_Full_DB}
}

@online{shamela,
  title = {al-Maktaba al-Shamela},
  url = {https://shamela.ws}
}

---