theshivam7/ncert-dataset

📘 NCERT Educational Question-Answering Dataset

This dataset is a curated collection of question-answer pairs derived from NCERT (National Council of Educational Research and Training) textbooks used in Indian schools. It includes English and Science content for Classes 6, 7, and 8, structured to support NLP research in education, question answering, reading comprehension, and automated tutoring systems.

� Research & Code

• 📜 Paper: Pustak AI: Curriculum-Aligned and Interactive Textbooks Using Large Language Models
• 💻 GitHub Repository: Pustak-AI-artifacts

�👥 Authors

• Shivam Sharma (IIT Madras)
• Riya Naik (BITS Pilani)
• Tejas Gawas
• Heramb Patil
• Prof. Kunal Kargaonkar (BITS Pilani)

🧾 Dataset Summary

Dataset	Subject	Class	Questions	Chapters	Avg Questions/Chapter
ncert-class6-english	English	6	~144	8	~18
ncert-class7-english	English	7	~181	8	~23
ncert-class8-english	English	8	~125	8	~16
ncert-class6-8-english-combined	English	6–8	~450	24	~19
ncert-class6-science	Science	6	~77	13	~6
ncert-class7-science	Science	7	~108	13	~8
ncert-class8-science	Science	8	~102	13	~7
ncert-class6-8-science-combined	Science	6–8	~287	39	~7
ncert-class6-social-science	Social Science	6	~64	7	~9

🧬 Dataset Structure

Each record includes:

{
  "context": "The Harappan civilization was one of the earliest urban societies in the world...",
  "question": "What is the Harappan civilization known for?",
  "answer": "It is known for being one of the earliest urban civilizations."
}

📊 Data Splits

All configs expose a single train split containing the full dataset. There are no pre-defined validation or test splits. For fine-tuning experiments, we recommend an 80/10/10 stratified split by question type.

🛠️ Data Curation Process

1. Text Extraction from NCERT PDF textbooks using custom scripts.
2. Manual QA Curation to ensure pedagogical quality.
3. Contextual Mapping to maintain alignment between questions and source text.
4. Formatting in CSV files with standard columns: context, question, answer.
5. Validation & Deduplication for quality and uniqueness.

---

🔧 Usage Example

You can load the dataset using the Hugging Face Datasets library:

Install 🤗 Datasets:

pip install datasets

Load via:

from datasets import load_dataset

# English — Classes 6–8 combined (default)
ds = load_dataset("theshivam7/ncert-dataset", name="english")

# English — individual classes
ds = load_dataset("theshivam7/ncert-dataset", name="english_class6")
ds = load_dataset("theshivam7/ncert-dataset", name="english_class7")
ds = load_dataset("theshivam7/ncert-dataset", name="english_class8")

# Science — Classes 6–8 combined
ds = load_dataset("theshivam7/ncert-dataset", name="science")

# Science — individual classes
ds = load_dataset("theshivam7/ncert-dataset", name="science_class6")
ds = load_dataset("theshivam7/ncert-dataset", name="science_class7")
ds = load_dataset("theshivam7/ncert-dataset", name="science_class8")

# Social Science — Class 6
ds = load_dataset("theshivam7/ncert-dataset", name="social_science")

Potential Use Cases

1. 📖 Reading Comprehension Modeling
2. 🧠 Curriculum-Aware Tutoring Systems
3. ✍️ Automated Question Generation
4. 📊 Curriculum & Assessment Analytics
5. 📚 Content Summarization
6. 🌐 Language Learning Tools for Indian Classrooms

License

This dataset is released under the Creative Commons Attribution 4.0 (CC BY 4.0) license. See the LICENSE file for more information.

Citation

If you use this dataset in your research or application, please cite our paper:

@article{pustakai2025,
  title={Pustak AI: Curriculum-Aligned and Interactive Textbooks Using Large Language Models},
  author={Shivam Sharma and Riya Naik and Tejas Gawas and Heramb Patil and Kunal Kargaonkar},
  journal={arXiv preprint arXiv:2511.10002v2},
  year={2025}
}

---