|
---
|
|
language:
|
|
- en
|
|
- ig
|
|
pretty_name: "My English to Igbo Dataset"
|
|
license: mit
|
|
tags:
|
|
- machine-translation
|
|
- igbo
|
|
- english
|
|
- nlp
|
|
---
|
|
|
|
# English-Igbo Parallel Corpus
|
|
|
|
## Description
|
|
|
|
This dataset is a comprehensive collection of parallel sentences in English and Igbo. It has been compiled from multiple sources to create a rich resource for machine translation, language research, and natural language processing tasks. The dataset is particularly valuable for those focusing on Igbo, a language spoken primarily in Nigeria, which is underrepresented in the field of computational linguistics.
|
|
|
|
## Composition
|
|
|
|
The dataset comprises the following components:
|
|
|
|
1. **iamwille/igbo-translation Dataset:** Originally sourced from [Hugging Face](https://huggingface.co/datasets/iamwille/igbo-translation), this subset includes professionally translated sentences covering various topics.
|
|
|
|
2. **Igbo-English Machine Translation Dataset:** This subset, also sourced from [Papers with Code](https://paperswithcode.com/dataset/igbonlp-datasets), provides additional translated sentences with a focus on common phrases and expressions.
|
|
|
|
3. **Custom Text File Translations:** A set of [English-Igbo](https://github.com/masakhane-io/masakhane-community/blob/master/list-of-datasets.md) sentence pairs extracted from a custom text file, offering a diverse range of everyday language usage.
|
|
|
|
4. **JW300 Text Data:** Sourced from the [JW300 corpus](https://github.com/Niger-Volta-LTI/igbo-text/tree/master/JW300), this part includes religious and educational text, contributing to the diversity of the dataset.
|
|
|
|
## Dataset Structure
|
|
|
|
### Data Splits
|
|
|
|
The dataset is divided into two splits:
|
|
|
|
- **Training Set:** Contains a majority of the sentences, intended for training machine translation models or other NLP tasks.
|
|
- **Test Set:** A smaller set of sentences for evaluating model performance.
|
|
|
|
### Data Fields
|
|
|
|
Each entry in the dataset consists of two fields:
|
|
|
|
- `English`: The sentence or phrase in English.
|
|
- `Igbo`: The corresponding translation in Igbo.
|
|
|
|
### Data Format
|
|
|
|
The dataset is available in CSV format, with each row representing a parallel sentence pair.
|
|
|
|
## Usage
|
|
|
|
This dataset can be used for various tasks in NLP, including but not limited to:
|
|
|
|
- Machine Translation
|
|
- Cross-lingual Transfer Learning
|
|
- Linguistic Research
|
|
|
|
## Source and Acknowledgments
|
|
|
|
This dataset has been compiled from various sources. We acknowledge the contributions of:
|
|
|
|
- Hugging Face for hosting the initial datasets.
|
|
- The creators of the JW300 corpus.
|
|
- Individual contributors of the custom text translations.
|
|
|
|
## Licensing
|
|
|
|
This dataset is released under [specify the license, e.g., MIT License]. Please ensure to adhere to the licensing terms of the original sources when using this dataset.
|
|
|
|
## Contact
|
|
|
|
For any queries or contributions, please contact Chinemerem Ibe-Ekeocha at ccibeekeoc42@gmail.com.
|
|
|