README.md

metadata

dataset_info:
  features:
    - name: french
      dtype: string
    - name: moore
      dtype: string
    - name: source
      dtype: string
    - name: source_lang
      dtype: string
    - name: target_lang
      dtype: string
    - name: id
      dtype: int64
    - name: original_id
      dtype: int64
    - name: was_corrected
      dtype: bool
    - name: quality_warnings
      dtype: string
    - name: len_ratio
      dtype: float64
  splits:
    - name: train
      num_bytes: 33741869
      num_examples: 119465
    - name: val
      num_bytes: 445026
      num_examples: 1437
    - name: test
      num_bytes: 900925
      num_examples: 2931
  download_size: 16480767
  dataset_size: 35087820
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: val
        path: data/val-*
      - split: test
        path: data/test-*

Superseded by burkimbia/moore-parallel-corpus, which publishes every release as a tag of one repository (first tag v0.8.0, from bia-datasets-text v0.8.0). Its val/test are not comparable with this version's: pairs are now placed by a hash of their normalized text (ADR-016). It also carries a defect fixed later: 12 098 Bible JW French verses lost a comma or period to a control byte (\x01). This repository is kept unchanged so that earlier results stay reproducible.