Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,57 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc-by-nc-4.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-nc-4.0
|
| 3 |
+
---
|
| 4 |
+
# ArQuAD: An Expert-Annotated Arabic Machine Reading Comprehension Dataset
|
| 5 |
+
|
| 6 |
+
## Overview
|
| 7 |
+
|
| 8 |
+
ArQuAD is an expert-annotated Arabic Machine Reading Comprehension (MRC) dataset. It comprises 16,020 questions posed by language experts on passages extracted from the most frequently visited Arabic Wikipedia articles. Each question's answer is a text segment from the corresponding reading passage.
|
| 9 |
+
|
| 10 |
+
## Citation
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
If you use ArQuAD in your research, please cite the following paper:
|
| 14 |
+
|
| 15 |
+
bibtex
|
| 16 |
+
@article{obeidat2024arquad,
|
| 17 |
+
title={ArQuAD: An Expert-Annotated Arabic Machine Reading Comprehension Dataset},
|
| 18 |
+
author={Obeidat, Rasha and Al-Harbi, Marwa and Al-Ayyoub, Mahmoud and Alawneh, Luay},
|
| 19 |
+
journal={Cognitive Computation},
|
| 20 |
+
pages={1--20},
|
| 21 |
+
year={2024},
|
| 22 |
+
publisher={Springer}
|
| 23 |
+
}
|
| 24 |
+
## Dataset Description
|
| 25 |
+
|
| 26 |
+
ArQuAD consists of 16,020 question-answer pairs created by Arabic language specialists with BA and MA degrees. The passages are sourced from 1335 of the most viewed Arabic Wikipedia articles, covering a wide range of topics including sports, politics, technology, religion, and more.
|
| 27 |
+
|
| 28 |
+
### Structure
|
| 29 |
+
|
| 30 |
+
The dataset is provided in both CSV and SQuAD JSON formats with the following columns:
|
| 31 |
+
- `passage`: The original passage from Wikipedia.
|
| 32 |
+
- `question`: The question posed by the annotator.
|
| 33 |
+
- `answer`: The minimal text span from the passage that answers the question.
|
| 34 |
+
|
| 35 |
+
### Statistics
|
| 36 |
+
|
| 37 |
+
The dataset includes:
|
| 38 |
+
- Total pairs: 16,020
|
| 39 |
+
- Passages: 4,005
|
| 40 |
+
- Domains covered: Various (sports, politics, technology, etc.)
|
| 41 |
+
|
| 42 |
+
### Key Features
|
| 43 |
+
|
| 44 |
+
- **Expert-Annotated**: Questions and answers are created by language experts, ensuring high quality and relevance.
|
| 45 |
+
- **Diverse**: Covers a wide range of topics to ensure comprehensive testing of MRC models, also includes a mix of factoid and non-factoid questions.
|
| 46 |
+
|
| 47 |
+
## Usage
|
| 48 |
+
|
| 49 |
+
Download the CSV file from the repository to use this dataset and load it into your preferred data analysis tool.
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
## Contact
|
| 53 |
+
|
| 54 |
+
or any questions or issues regarding the dataset, please contact:
|
| 55 |
+
Rasha Obeidat:rmobeidat@just.edu.jo (or any of the authors)
|
| 56 |
+
|
| 57 |
+
|