DarwinDanish/bahasa-iban: Bahasa Iban Text Corpus

📝 Dataset Description

The Bahasa Iban Text Corpus is a collection of monolingual text data in Bahasa Iban, an indigenous language primarily spoken by the Iban people of Sarawak, Malaysia, and parts of Brunei and Indonesia.

This dataset is specifically curated to support Text Generation tasks and general Natural Language Processing (NLP) research for this low-resource language. Its primary goal is to provide a foundational corpus for training large language models (LLMs), language models, and various NLP components, thereby contributing to the digital preservation and technological advancement of the Iban language.

The corpus size falls within the 1M to 10M size category.

💾 Dataset Structure

This dataset is structured as a monolingual corpus. It is designed to be easily loaded and processed for language modeling.

The dataset contains the following split:

Split	Number of Examples (Approx.)
train	1.8 million

Data Fields

The data points consist of a single feature:

Feature Name	Data Type	Description
text	string	The actual text or document segment in Bahasa Iban.

💻 How to Use the Dataset

To load the bahasa-iban dataset using the Hugging Face datasets library, use the following code snippet:

from datasets import load_dataset

# Load the main split of the dataset
dataset = load_dataset("DarwinDanish/bahasa-iban", split="train")

# Access the first example
print(dataset[0])

# Example output:
# {'text': 'Sida iya lalu bejalai ngagai rumah panjai ti endur alai sida iya belelak ke sekejap laban sida iya udah lelak amai bejalai. . .'}

🧑‍💻 Dataset Creation

Curation Rationale This dataset was created to address the significant scarcity of publicly available, centralized text data for Bahasa Iban. By aggregating text from various sources, this corpus aims to lower the barrier to entry for researchers and developers working on Iban NLP, especially for tasks like fine-tuning generative models. Source Data The data in this corpus was collected from a variety of publicly accessible sources written in Bahasa Iban. These sources may include, but are not limited to:

• Digitized folk tales and oral histories.
• Online news articles and publications.
• Social media content or public discussions. Curator The dataset was created and curated by DarwinDanish.

⚖️ Licensing Information

The use of this dataset is governed by an explicit non-standard license:

• License Name: lumina-license
• License Link: LICENSE Users must refer to the LICENSE file in the repository for full legal terms and conditions of use.

⚠️ Limitations and Biases

• Domain Specificity: The text sources may introduce biases towards certain topics (e.g., news, cultural narratives) depending on the source material used for collection.
• Low Resource Challenges: As a low-resource language, the corpus size, while substantial, may not fully cover the diversity of the language, dialects, and grammatical structures found in all Iban-speaking communities.
• Data Cleaning: While efforts were made to clean the data, residual noise, spelling variations, or inconsistent orthography inherent to low-resource language data may still be present.

📜 Citation

If you use this dataset in your work, please consider citing it appropriately.

@misc{darwindanish_bahasaiban_2024,
  author = {DarwinDanish},
  title = {Bahasa Iban Text Corpus for Text Generation},
  year = {2024},
  publisher = {Hugging Face},
  url = {[https://huggingface.co/datasets/DarwinDanish/bahasa-iban](https://huggingface.co/datasets/DarwinDanish/bahasa-iban)}
}