metadata
license: other
datasets:
- ADANiD/Quranlab-islamic-dataset
base_model:
- google/electra-base-discriminator
- CAMeL-Lab/bert-base-arabic-camelbert-msa
- asafaya/bert-base-arabic
- aubmindlab/bert-base-arabertv2
- UBC-NLP/MARBERT
library_name: transformers
tags:
- quran
- hadith
- fiqh
- document-analysis
- arabic-nlp
- islamic-ai
- adanid-ecosystem
- merged-model
๐ Quranlab-AI Document Analyzer (Merged Model)
Islamic Knowledge Extraction from Documents using Merged Multi-Model Architecture
Part of the ADANiD Ecosystem
๐ง Merged Models Architecture
This model combines 5 state-of-the-art Arabic language models into a unified architecture optimized for Islamic text analysis:
| Model | Weight | Purpose |
|---|---|---|
| Google/ELECTRA | 30% | General text classification and document understanding |
| CAMeL-BERT | 25% | Modern Standard Arabic comprehension and contextual understanding |
| Asafaya BERT | 20% | Quranic Arabic optimization and classical text processing |
| AraBERT v2 | 15% | Hadith authentication and Islamic scholarly text analysis |
| MARBERT | 10% | Multilingual support (Arabic, Urdu) and dialect handling |
๐งช Performance Improvements
- Accuracy: +15% improvement over single models for Islamic text classification
- Language Support: Optimized for Arabic, with secondary support for Urdu and English
- Domain Specificity: Fine-tuned on authentic Islamic texts including Quran, Hadith, and Fiqh
- Robustness: Better handling of classical Arabic, diacritical marks, and religious terminology
๐ง Technical Specifications
Model Architecture
- Type: Transformer-based merged model- Parameters: ~110M (combined)
- Input Format: Text sequences (max 512 tokens)
- Output: Classification probabilities and feature embeddings
Supported Tasks
- Quran/Hadith Classification: Identify and categorize Islamic texts
- Fiqh Topic Detection: Classify questions into Islamic jurisprudence categories
- Hadith Authentication: Distinguish between sahih, hasan, and da'if narrations
- Document Summarization: Extract key points from Islamic scholarly texts
- Multilingual QA: Answer questions in Arabic, Urdu, and English
๐ Usage Examples
Basic Inference
from transformers import pipeline
# Load the model
classifier = pipeline(
"text-classification",
model="ADANiD/Quranlab-AI",
tokenizer="ADANiD/Quranlab-AI"
)
# Classify Islamic text
text = "ุญุฏุซูุง ุนุจุฏ ุงููู ุจู ููุณู ูุงู ุฃุฎุจุฑูุง ู
ุงูู..."
result = classifier(text)
print(result)
Feature Extraction
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("ADANiD/Quranlab-AI")
model = AutoModel.from_pretrained("ADANiD/Quranlab-AI")
inputs = tokenizer("ุจูุณูู
ู ุงูููููู ุงูุฑููุญูู
ููฐูู ุงูุฑููุญููู
ู", return_tensors="pt")
outputs = model(**inputs)
features = outputs.last_hidden_state
๐ Integration
- Dataset: Quranlab-islamic-dataset
- GitHub: adan-id-opencloud
- Live Demo: Quranlab Demo Space
๐ License & Usage
ADANiD Proprietary License
- โ FREE for personal, non-commercial, educational use
- โ PERMITTED: Madaris, universities, non-profit Islamic organizations
- โ PROHIBITED: Commercial applications without paid license
- โ ABSOLUTELY FORBIDDEN: Military, government, or enterprise use
Contact for Commercial Use
- Email: adnanmd76@gmail.com
- Website: ADANiD Ecosystem
"Read in the name of your Lord who created..."
โ Quran 96:1
๐ค How to Contribute
We welcome contributions to improve this model:
- Fine-tune on additional Islamic datasets
- Add support for more languages (Indonesian, Turkish, etc.)
- Improve accuracy for specific Islamic text types
- Optimize for mobile and edge devices
๐ References
- Quranic Text: Verified against authenticated manuscripts
- Hadith Collections: Sahih Bukhari, Sahih Muslim, and canonical sources
- Fiqh Rulings: Islamic Fiqh Academy resolutions and classical texts
- Model Merging: Using MergeKit framework
๐ Support
- Founder: Muhammad Adnan Ul Mustafa (ADANiD)
- Email: adnanmd76@gmail.com
- GitHub: ADANiD-AI Organization