Datasets:
Tasks:
Question Answering
Modalities:
Text
Formats:
json
Languages:
English
Size:
10K - 100K
ArXiv:
License:
๐ LCQA-Islamic: A Benchmark Dataset with Larger Context for Non-Factoid QA over Islamic Texts
Dataset Summary
This dataset provides a benchmark for non-factoid question answering over Islamic texts with an emphasis on larger context retrieval.
It includes expert-curated QA pairs where the answers require reasoning across multi-sentence or paragraph-level contexts from authentic Islamic sources such as the Quran, Hadith, and Tafseer.
Designed to support long-contextual Answer generation models and multi-passage reasoning tasks.
Supported Tasks and Leaderboards
- ๐ Long-Form Question Answering
- ๐ Multi-Passage Retrieval
- ๐ค Suitable for evaluation of RAG systems and long-context LLMs
Languages
- ๐ English
Dataset Structure
Each sample includes:
question
(string): The user querycontext
(string): Retrieved passages from Islamic textsanswer
(string): Ground-truth answer curated by experts
License
This dataset is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.
- ๐ Non-commercial use only
- โ๏ธ Attribution required
- ๐ ShareAlike required
Full license text: CC BY-NC-SA 4.0
Citation
If you use this dataset in your research, please cite the following paper:
@article{qamar2024benchmark,
title={A Benchmark Dataset with Larger Context for Non-Factoid Question Answering over Islamic Text},
author={Qamar, Faiza and Latif, Seemab and Latif, Rabia},
journal={arXiv preprint arXiv:2409.09844},
year={2024}
}
- Downloads last month
- 48