Datasets:

bot-yaya
/

rework_undl_text

configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
dataset_info:
  features:
    - name: ar
      dtype: string
    - name: zh
      dtype: string
    - name: en
      dtype: string
    - name: fr
      dtype: string
    - name: ru
      dtype: string
    - name: es
      dtype: string
    - name: de
      dtype: string
    - name: record
      dtype: string
  splits:
    - name: train
      num_bytes: 48622457871
      num_examples: 165840
  download_size: 3906189450
  dataset_size: 48622457871
license: mit
task_categories:
  - translation
language:
  - ar
  - en
  - fr
  - es
  - zh
  - ru
  - de
tags:
  - UnitedNations
  - UN
pretty_name: >-
  Parallel Corpus from United Nations Digital Library Official Document System
  aligned by file

Please, visit our Github repo at v4dev branch for up-to-date deploying notes and reproducing details. https://github.com/mnbvc-parallel-corpus-team/UPRPRC/tree/v4dev

The main branch will stay unchanged until review process is done.

2025/12/03 Updated

Hello everyone, please download the updated and more complete file-level alignment dataset: https://huggingface.co/datasets/bot-yaya/UPRPRC_FTXT_FILEWISE

Or, if you can't feed in the entire file, download the paragraph-level alignment dataset: https://huggingface.co/datasets/bot-yaya/UPRPRC_BLOCKWISE

This dataset is from an early version of UPRPRC and is incomplete. It's kept for reviewing submitted papers. The two datasets above are supersets of this dataset, including all the files within it.

Parallel Corpus from United Nations Digital Library ODS (2000-2023)

Data source link (the website logic has been restructured and updated since when this data was crawled, so inconsistencies may exist): https://search.un.org/search?collection=ods&currentPageNumber=1&q=*&row=10&sort=relevance

The source text from pandoc to docx was generated using the command: pandoc -i {filepath} -t plain -o {outpath} --strip-comments

This text may still require some noise reduction steps, such as removing separators consisting entirely of horizontal lines and table elements, before it can be used for subsequent translation and alignment steps.

Older data link: https://huggingface.co/datasets/bot-yaya/undl_text

Due to incorrect parameters in the old version, some data was lost during processing. Therefore, a new version has been created and re-uploaded. It is recommended to download and use this new version instead of the old one.

2025/12/03 Updated

@全体成员都给我去下更新更全的文件级对齐数据集：https://huggingface.co/datasets/bot-yaya/UPRPRC_FTXT_FILEWISE

或者你喂不下整份文件的话去下段落级对齐的：https://huggingface.co/datasets/bot-yaya/UPRPRC_BLOCKWISE

这份数据集是早期版本的UPRPRC产物，东西不全，现在留着没删是为了已经投出去的论文review用。上面两个数据集是这个数据集的超集，包括了这份里面所有的文件

联合国数字图书馆ODS里爬出来的平行语料 Parallel Corpus from United Nations Digital Library ODS（2000-2023）

数据源链接（网站逻辑比起爬取这些数据时已经重构更新，可能会有不一致的情况）：https://search.un.org/search?collection=ods&currentPageNumber=1&q=*&row=10&sort=relevance

pandoc转docx出的源文本，所用命令为：pandoc -i {filepath} -t plain -o {outpath} --strip-comments

这些文本可能仍需一定的步骤去噪，比如去掉全是横线的分隔符、去掉表格元素，才能用于后续的翻译及对齐步骤

旧版数据链接 https://huggingface.co/datasets/bot-yaya/undl_text

因为旧版参数不当，处理的时候丢掉了一部分数据，所以重做了一份重新上传，建议是下载使用这份，而不是旧版