oliverkinch/machine-translation-da-ar

---
language:
- da
- ar
license: cc0-1.0
task_categories:
- translation
tags:
- danish
- arabic
- opus
- parallel-corpus
- translation
pretty_name: Danish-Arabic Translation Dataset
size_categories:
- 1M<n<10M
source_datasets:
- opus
---

# Danish-Arabic Machine Translation Dataset

A Danish–Arabic parallel corpus assembled from multiple OPUS sources and filtered using heuristic quality rules.

## Dataset Summary

| Attribute | Value |
| ------------------------------- | ---------------------------- |
| Columns | `danish`, `arabic`, `source` |
| Language pair | Danish → Arabic |
| Sentence length range (Danish) | 4 – 150 words |

## Load the dataset

```python
from datasets import load_dataset

ds = load_dataset("oliverkinch/machine-translation-da-ar", split="train")
print(ds[0])
# {'danish': '...', 'arabic': '...', 'source': 'NLLB'}
```

## Data Fields

| Field | Type | Description |
| --------- | ------ | -------------------------- |
| `danish` | string | Source sentence in Danish |
| `arabic` | string | Target sentence in Arabic |
| `source` | string | Originating corpus name |

## Source Corpora

The dataset aggregates sentence pairs from the following OPUS corpora:

| Source | Domain |
| ---------------- | ------------ |
| WikiMatrix | Wikipedia |
| TED2020 | TED Talks |
| ELRC-wikipedia_health | Health |
| wikimedia | Wikipedia |
| NLLB | Web |

WikiMatrix and wikimedia align Wikipedia content across languages. TED2020 provides parallel transcripts of TED talks. ELRC-wikipedia_health covers health and COVID-19 related Wikipedia content. NLLB (No Language Left Behind) is a multilingual corpus released by Meta AI.

## Filtering Pipeline

Pairs were retained only if they passed all of the following heuristic filters:

| Filter | Criterion |
| ----------------------------- | --------- |
| Minimum words (Danish) | ≥ 4 |
| Minimum words (Arabic) | ≥ 3 |
| Maximum words (either side) | ≤ 150 |
| DA/AR word-count ratio | 0.4 – 2.5 |
| URLs in either field | Rejected |
| Non-Latin script in Danish | Rejected |
| Non-Arabic script in Arabic | Rejected |
| Exact deduplication (Danish) | Enabled |

## Intended Uses

* Training and fine-tuning machine translation (MT) models for the Danish–Arabic direction
* Data augmentation for low-resource Danish and Arabic NLP tasks
* Benchmarking cross-lingual representations

## Licensing

Individual source corpora carry their own licenses (primarily CC-BY-4.0 and CC-BY-SA-4.0). Please consult the [OPUS platform](https://opus.nlpl.eu) for the license of each sub-corpus before use in commercial settings.

## Citation

If you use this dataset, please cite the relevant OPUS sub-corpora. The OPUS platform is described in:

```bibtex
@inproceedings{tiedemann-2012-parallel,
title = {Parallel Data, Tools and Interfaces in {OPUS}},
author = {Tiedemann, J{\"o}rg},
booktitle = {Proceedings of the Eight International Conference on Language Resources and Evaluation ({LREC}'12)},
year = {2012},
pages = {2214--2218}
}
```