Datasets:

Davidsamuel101
/

ebible_local_ind_corpus

language:
  - ind
  - aaz
  - heg
  - lex
  - llg
  - nfa
  - ptu
  - rgu
  - row
  - tet
  - txq
  - wrs
task_categories:
  - translation
pretty_name: eBible Indonesian Local Language Corpus
size_categories:
  - 10K<n<100K
configs:
  - config_name: ind_aaz
    data_files:
      - split: train
        path: ind_aaz/train-*
  - config_name: ind_heg
    data_files:
      - split: train
        path: ind_heg/train-*
  - config_name: ind_lex
    data_files:
      - split: train
        path: ind_lex/train-*
  - config_name: ind_llg
    data_files:
      - split: train
        path: ind_llg/train-*
  - config_name: ind_nfa
    data_files:
      - split: train
        path: ind_nfa/train-*
  - config_name: ind_ptu
    data_files:
      - split: train
        path: ind_ptu/train-*
  - config_name: ind_rgu
    data_files:
      - split: train
        path: ind_rgu/train-*
  - config_name: ind_row
    data_files:
      - split: train
        path: ind_row/train-*
  - config_name: ind_tet
    data_files:
      - split: train
        path: ind_tet/train-*
  - config_name: ind_txq
    data_files:
      - split: train
        path: ind_txq/train-*
  - config_name: ind_wrs
    data_files:
      - split: train
        path: ind_wrs/train-*
dataset_info:
  - config_name: default
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
  - config_name: ind_heg
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
    splits:
      - name: train
        num_bytes: 3426289
        num_examples: 9073
    download_size: 1664016
    dataset_size: 3426289
  - config_name: ind_lex
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
    splits:
      - name: train
        num_bytes: 3623717
        num_examples: 9674
    download_size: 1804057
    dataset_size: 3623717
  - config_name: ind_llg
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
    splits:
      - name: train
        num_bytes: 3490656
        num_examples: 9071
    download_size: 1664187
    dataset_size: 3490656
  - config_name: ind_nfa
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
    splits:
      - name: train
        num_bytes: 3595249
        num_examples: 9070
    download_size: 1658271
    dataset_size: 3595249
  - config_name: ind_ptu
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
    splits:
      - name: train
        num_bytes: 3222522
        num_examples: 9345
    download_size: 1555084
    dataset_size: 3222522
  - config_name: ind_rgu
    features:
      - name: source_text
        dtype: string
      - name: target_text
        dtype: string
      - name: source_lang
        dtype: string
      - name: target_lang
        dtype: string
      - name: verse
        dtype: string
    splits:
      - name: train
        num_bytes: 3197760
        num_examples: 8223
    download_size: 1515957
    dataset_size: 3197760

eBible Indonesian Local Language Corpus

This dataset contains parallel Bible translations between Indonesian and various local languages from Indonesia, particularly from Eastern Indonesia regions.

Dataset Description

This dataset is created from the eBible corpus, containing verse-aligned translations between Indonesian and multiple local languages. The dataset properly handles verse ranges where multiple verses are combined in the translation.

Available Language Pairs

The dataset contains the following language pairs as different configurations/subsets:

Configuration	Source	Target	Language Name	Examples	Description
`ind_aaz`	ind	aaz	Amarasi	9,068	Amarasi Bible translation
`ind_heg`	ind	heg	Helong	9,073	Helong Bible translation
`ind_lex`	ind	lex	Luang	9,674	Luang Bible translation
`ind_llg`	ind	llg	Lole	9,071	Lole Bible translation
`ind_nfa`	ind	nfa	Dhao	9,070	Dhao Bible translation
`ind_ptu`	ind	ptu	Bambam	9,345	Bambam Bible translation
`ind_rgu`	ind	rgu	Rikou	8,223	Rikou Bible translation
`ind_row`	ind	row	Dela-Oenale	9,070	Dela-Oenale Bible translation
`ind_tet`	ind	tet	Tetun	9,069	Tetun Bible translation
`ind_txq`	ind	txq	Tii	9,069	Tii Bible translation
`ind_wrs`	ind	wrs	Waris	8,981	Waris Bible translation

Total examples across all configurations: 99,713

Dataset Structure

Each configuration contains a 'train' split with the following features:

source_text: Text in the source language (Indonesian)
target_text: Text in the target language
source_lang: Source language code (always 'ind')
target_lang: Target language code
verse: Bible verse reference (e.g., "GEN 1:1" or "GEN 1:14-15" for verse ranges)

Usage

Loading a specific language pair

from datasets import load_dataset

# Load a specific language pair (configuration)
dataset = load_dataset("Davidsamuel101/ebible_local_ind_corpus", "ind_aaz")
train_data = dataset['train']

# Example: access first example
print(train_data[0])

Loading multiple configurations

from datasets import load_dataset

# Load specific configurations
target_langs = ["aaz", "ptu", "nfa", "heg", "lex", "row", "llg", "rgu", "txq", "tet", "wrs"]
for lang in target_langs:
    dataset = load_dataset("Davidsamuel101/ebible_local_ind_corpus", f"ind_{lang}")
    print(f"{lang}: {len(dataset['train'])} examples")

Using for machine translation

from datasets import load_dataset
from transformers import pipeline

# Load Indonesian to Tetun dataset
dataset = load_dataset("Davidsamuel101/ebible_local_ind_corpus", "ind_tet")['train']

# Example: prepare for training
def preprocess_function(examples):
    inputs = [ex for ex in examples["source_text"]]
    targets = [ex for ex in examples["target_text"]]
    return {"translation": [{"ind": i, "tet": t} for i, t in zip(inputs, targets)]}

# Apply preprocessing
processed_dataset = dataset.map(preprocess_function, batched=True)

Data Processing Notes

The dataset was created with special handling for:

Verse ranges: Marked with <range> tokens in the original corpus
Empty verses: Automatic skipping of verses where either source or target text is blank
Verse concatenation: Proper concatenation of source verses when targets combine multiple verses

Languages

Source Language

Indonesian (ind): Bahasa Indonesia, the national language of Indonesia

Target Languages

Amarasi (aaz): Uab Meto language spoken in West Timor
Helong (heg): Austronesian language of West Timor
Luang (lex): Language spoken in the Luang islands
Lole (llg): Language of Central Rote
Dhao (nfa): Language of Ndao island
Bambam (ptu): Language of Central Sulawesi
Rikou (rgu): Language of Central Rote
Dela-Oenale (row): Language of West Rote
Tetun (tet): Language of East Timor
Tii (txq): Language of West Rote
Waris (wrs): Papuan language

Citation

If you use this dataset, please cite:

@misc{ebible_indonesian_2024,
  title={eBible Indonesian Local Language Corpus},
  author={Dataset created from eBible corpus},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/Davidsamuel101/ebible_local_ind_corpus}
}

License

Please refer to the original eBible corpus license terms. The texts are from Bible translations which may have their own usage terms.

Acknowledgments

This dataset is derived from the eBible corpus. We acknowledge the translators and communities who created these Bible translations in local languages.