large-traversaal/openbookqa_urdu_final

# Dataset Card: OpenBookQA Urdu

## Dataset Summary

`openbookqa_urdu_cleaned` is a cleaned Urdu translation of the **OpenBookQA** dataset, a multiple-choice question answering benchmark designed to test **elementary science understanding combined with commonsense reasoning**. Each example consists of a question and four answer options, with exactly one correct answer.

The dataset provides Urdu translations of questions and answer choices, enabling evaluation and training of **Urdu and multilingual language models** on scientific reasoning tasks in a low-resource language setting.

## Dataset Details

* **Dataset Name:** openbookqa_urdu_cleaned
* **Maintained by:** large-traversaal (Traversaal.ai)
* **Task Type:** Multiple-choice question answering
* **Domain:** Elementary science and commonsense reasoning
* **Languages:** Urdu (primary), English (where original fields are retained)
* **Format:** Parquet
* **Answer Choices:** 4 per question

## Dataset Structure

Each record typically contains the following fields:

* `id`: Unique example identifier
* `question`: Urdu translation of the question
* `choices`: Answer options (four choices, labeled A–D)
* `answerKey`: Correct answer label (A, B, C, or D)
* `english_question` (optional): Original English question
* `english_choices` (optional): Original English answer options

Exact field names may vary slightly depending on split and preprocessing version.

## Intended Uses

This dataset is intended for:

* Training and evaluating Urdu and multilingual QA models
* Benchmarking reasoning performance on science-based questions
* Cross-lingual transfer learning from English to Urdu
* Research in low-resource language understanding and reasoning

## Loading the Dataset
```python
from datasets import load_dataset

ds = load_dataset("large-traversaal/openbookqa_urdu_cleaned")
```

## Licensing and Usage

Licensing follows the terms of the original OpenBookQA dataset. Users should verify license details on the Hugging Face dataset page before redistribution or commercial use.

## 📄 Citation

If you use this dataset in your research, please cite the **UrduBench paper**:

```bibtex
@misc{shafique2026urdubenchurdureasoningbenchmark,
title={UrduBench: An Urdu Reasoning Benchmark using Contextually Ensembled Translations with Human-in-the-Loop},
author={Muhammad Ali Shafique and Areej Mehboob and Layba Fiaz and Muhammad Usman Qadeer and Hamza Farooq},
year={2026},
eprint={2601.21000},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2601.21000}
}
```