beethogedeon/Helsinki-en-fon-opus-20k

English-Fon Translation Dataset

Dataset Description

This dataset contains parallel English-Fon (Fongbe) translation pairs, created by translating English texts from the Helsinki-NLP/opus-100 dataset (en-fr subset) into Fon using Google Translate via the deep_translator Python library.

Fon (Fongbe) is a Gbe language spoken primarily in Benin by approximately 8 million people. It is one of the major languages of Benin.

Dataset Structure

Data Fields

Column	Type	Description
source_text	string	The original English text
source_lang	string	Source language code (en)
translated_text	string	The translated text in Fon
target_lang	string	Target language code (fon)

Data Splits

Split	Examples
train	90% of total
test	10% of total

Example

{
  "source_text": "The time now is 05:08.",
  "source_lang": "en",
  "translated_text": "Ganxixo dìn wɛ nyí 05:08.",
  "target_lang": "fon"
}

Usage

from datasets import load_dataset

dataset = load_dataset("beethogedeon/Helsinki-en-fon-opus-20k")

# Access train split
train = dataset["train"]
print(train[0])

# Use for training a translation model
for example in train:
    source = example["source_text"]
    target = example["translated_text"]

Dataset Creation

Source Data

The English source texts were extracted from the en-fr subset of Helsinki-NLP/opus-100, which is a collection of parallel corpora covering 100 language pairs from the OPUS collection.

Translation Process

• Tool: deep_translator Python library (GoogleTranslator backend)
• Source language: English (en)
• Target language: Fon (fon)
• Method: Automated machine translation with error handling and retry logic
• Post-processing: Translation errors and empty results were filtered out

Limitations and Biases

• Machine-translated: This dataset was created using Google Translate, which may produce errors, especially for a low-resource language like Fon. The translations have not been human-validated.
• Quality: Google Translate's quality for Fon is limited compared to high-resource languages. Some translations may be inaccurate, unnatural, or incomplete.
• Recommended use: This dataset can serve as a starting point for training translation models, but should ideally be refined through human review for production use.
• Source bias: The original OPUS-100 dataset may contain biases from its source corpora.

Intended Use

• Training and fine-tuning machine translation models for the English-Fon language pair
• Research on low-resource African language NLP
• Bootstrapping parallel corpora for Fon language processing
• Educational purposes for Fon language technology development

Citation

If you use this dataset, please cite the original OPUS-100 dataset:

@inproceedings{zhang2020improving,
  title={Improving Massively Multilingual Neural Machine Translation and Zero-Shot Translation},
  author={Zhang, Biao and Williams, Philip and Titov, Ivan and Sennrich, Rico},
  booktitle={Proceedings of ACL 2020},
  year={2020}
}

License

This dataset is released under CC-BY-4.0, consistent with the OPUS-100 source data licensing.

Contact

For questions, issues, or contributions (especially human-validated corrections), please open a discussion on this dataset's page.