---
dataset_info:
  features:
  - name: text
    dtype: string
  - name: sentences
    sequence: string
  - name: sentences_eng
    sequence:
      sequence: string
  splits:
  - name: train
    num_bytes: 467827355
    num_examples: 40367
  download_size: 260964625
  dataset_size: 467827355
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
language:
- tdt
- tet
pretty_name: "MADLAD-400 Tetun Dili translated to English"
---

# MADLAD-400-Tetun

The Tetun Dili "clean" split of the [MADLAD-400 dataset](https://huggingface.co/datasets/allenai/MADLAD-400/), sentencized and translated to English using [MADLAD-400 3b](https://huggingface.co/santhosh/madlad400-3b-ct2/).

Each row has: 
- `text`: the original text from MADLAD
- `sentences`: the text sentencized to Tetun, using [Moses](https://github.com/moses-smt/mosesdecoder), which has [Tetun non-breaking prefixes](https://github.com/moses-smt/mosesdecoder/blob/be3ae968baa7aa89cf17aa34957f1068f988aa9a/scripts/share/nonbreaking_prefixes/nonbreaking_prefix.tdt#L4).
- `sentences_eng`: the `sentences` translated to English, using MADLAD-400-3b and cTranslate2

Total of 884k sentences over 40k rows.