README.md

---
viewer: false
pretty_name: "Tigrinya-SQuAD: Machine-Translated Training Dataset"
language:
  - ti
multilinguality:
  - monolingual
task_categories:
  - question-answering
size_categories:
  - 10K<n<100K
dataset_size: ~10MB
download_size: ~6MB
license: cc-by-sa-4.0
tags:
  - tigrinya
  - question-answering
  - mrc
  - reading-comprehension
  - low-resource
  - african-languages
  - machine-translation
  - silver-standard
splits:
  - name: train
    num_examples: 46737
configs:
  - config_name: default
    data_files:
      - split: train
        path: "train.parquet"
---

# Tigrinya-SQuAD: Machine-Translated Training Dataset

Tigrinya-SQuAD is a machine-translated and filtered version of the English SQuAD 1.1 training dataset, automatically converted to Tigrinya for training question-answering models in low-resource settings.

This silver dataset serves as training data for Tigrinya question-answering systems. **For evaluation and benchmarking, please use the gold-standard [TiQuAD](https://huggingface.co/datasets/fgaim/tiquad) dataset, which contains human-annotated validation and test sets.**

**Published with the paper:** [Question-Answering in a Low-resourced Language: Benchmark Dataset and Models for Tigrinya](https://aclanthology.org/2023.acl-long.661/) (ACL 2023)

**Related repositories:**

- [TiQuAD (gold dataset)](https://huggingface.co/datasets/fgaim/tiquad)
- The paper's [GitHub repository](https://github.com/fgaim/TiQuAD)

## Dataset Overview

Tigrinya-SQuAD is designed as training data for extractive question answering in Tigrinya, a low-resource Semitic language primarily spoken in Eritrea and Ethiopia. The dataset features:

- **Source data**: English SQuAD 1.1 training part, which is based on Wikipedia articles
- **Machine-translated**: Automatically translated from English SQuAD 1.1 using neural machine translation
- **Filtered**: Post-processed with heuristic filtering to improve quality and discarded low-quality samples
- **Training-only**: Contains only training split; use TiQuAD for validation/testing
- **SQuAD format**: Maintains compatibility with standard QA frameworks
- Not human verified, to be used for training but not for final evaluation

| **Split** | **Articles** | **Paragraphs** | **Questions** | **Answers** |
|-----------|--------------|----------------|---------------|-------------|
| Train     | 442          | 17,391         | 46,737        | 46,737      |

## How to Load Tigrinya-SQuAD

Install the `datasets` library installed by running `pip install -U datasets` in the terminal.

> Make sure the latest `datasets` library is installed as older versions may not properly load the data.

Then pull and load the dataset using Python, as follows:

```python
from datasets import load_dataset

# Load the dataset
tigrinya_squad = load_dataset("fgaim/tigrinya-squad")
print(tigrinya_squad)
```

That will print the dataset features:

```python
DatasetDict({
    train: Dataset({
        features: ['id', 'question', 'context', 'answers', 'article_title', 'context_id'],
        num_rows: 46737
    })
})
```

### Data Fields

- **`id`**: Unique identifier for each question
- **`question`**: The question to be answered (in Tigrinya)
- **`context`**: The paragraph containing the answer (in Tigrinya)
- **`answers`**: A list of dictionaries of candidate answers, each entry containing:
  - `text`: An answer string (training data has one answer per question)
  - `answer_start`: A starting position of answer string in the context
- **`article_title`**: Title of the source article
- **`context_id`**: Unique identifier of the context in the data split

## Evaluation and Benchmarking

This dataset contains only training data, for proper evaluation of Tigrinya question-answering models use [TiQuAD](https://huggingface.co/datasets/fgaim/tiquad), which provides multireference, human-annotated validation/test splits. Both datasets can be combined during training for best results as reported in the paper.

## Citation

If you use this dataset in your work, please cite the original TiQuAD paper:

```bibtex
@inproceedings{gaim-etal-2023-tiquad,
    title = "Question-Answering in a Low-resourced Language: Benchmark Dataset and Models for {T}igrinya",
    author = "Fitsum Gaim and Wonsuk Yang and Hancheol Park and Jong C. Park",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.acl-long.661",
    pages = "11857--11870",
}
```

## Data Quality and Limitations

As a machine-translated dataset, Tigrinya-SQuAD has inherent limitations:

- **Translation errors**: Some questions/answers may have translation artifacts
- **Cultural adaptation**: Context may not perfectly align with Tigrinya cultural references
- Not suitable for model evaluation or human performance comparison but for training purpose only.

If you identify any issues with the dataset, please contact us at <fitsum.gaim@kaist.ac.kr>.

## Acknowledgments

This dataset builds upon the foundational work of the Stanford Question Answering Dataset (SQuAD) and the human-annotated TiQuAD dataset. We thank the original SQuAD creators for making their data freely available.

## License

This work is licensed under a [Creative Commons Attribution-ShareAlike 4.0 International License](https://creativecommons.org/licenses/by-sa/4.0/).

<a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/"><img alt="Creative Commons License" style="border-width:0" src="https://licensebuttons.net/l/by-sa/4.0/88x31.png" /></a>