Datasets:

tahrirchi
/

lutfiy

dataset_info:
  features:
    - name: src_sent
      dtype: string
    - name: tgt_sent
      dtype: string
    - name: src_lang
      dtype: string
    - name: tgt_lang
      dtype: string
    - name: source
      dtype: string
  splits:
    - name: train
      num_bytes: 10784658
      num_examples: 39994
  download_size: 5255037
  dataset_size: 10784658
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
language:
  - en
  - uz
  - uzs
pretty_name: lutfiy
size_categories:
  - 10K<n<100K
license: mit
task_categories:
  - translation
tags:
  - lutfiy
  - southern uzbek
  - afghani uzbek

Lutfiy: Southern Uzbek Parallel Corpus for Translation

This repository contains a parallel corpus for the Southern Uzbek language, developed as part of the research paper "Filling the Gap for Uzbek: Creating Translation Resources for Southern Uzbek".

Dataset Description

The Southern Uzbek Parallel Corpus is a collection of approximately 40,000 sentence pairs, designed to support machine translation tasks involving the Southern Uzbek language. It includes:

Northern Uzbek - Southern Uzbek (37,415 pairs)
English - Southern Uzbek (2,579 pairs)

Usage

This dataset is intended for training and evaluating machine translation models involving the Southern Uzbek language.

To load and use dataset, run this script:

from datasets import load_dataset

lutfiy_corpus = load_dataset("tahrirchi/lutfiy", split='train')

Dataset Structure

Data Instances

Size of downloaded dataset files: 5.3 MB
Size of the generated dataset: 4.9 MB
Total amount of disk used: 10.2 MB

An example from the corpus looks as follows.

{'src_sent': "Bolakay ko'kka boqib, to'rg'ay hazratning boshi ustida aylanib sayrayotganini ko'rdi.",
 'tgt_sent': 'بالهکهی کۉککه باقیب، تۉرغهی حرضتنینگ باشی اوستیده ایلهنیب َسیرهیاتگنینی کۉردی.',
 'src_lang': 'uzn_Latn',
 'tgt_lang': 'uzs_Arab',
 'source': 'books'}

Data Fields

The data fields are the same among all splits.

src_lang: a string feature that contains source language.
src_sent: a string feature that contains sentence in source language.
tgt_lang: a string feature that contains target language.
tgt_sent: a string feature that contains sentence in target language.
source: a string feature that contains source information of the pair.

Data Sources

The corpus comprises diverse parallel texts sourced from multiple domains:

90% sentences from books (novels, non-fiction)
6% sentences from internet websites
4% sentences from bilingual dictionaries

Data Preparation

The data mining process involved local mining techniques, ensuring that parallel sentences were extracted from translations of the same book, document, or article. Sentence alignment was performed using LaBSE (Language-agnostic BERT Sentence Embedding) embeddings. For more information, plase refet to our paper.

Citation

If you use this dataset in your research, please cite our paper:

@misc{mamasaidov2025fillinggapuzbekcreating,
      title={Filling the Gap for Uzbek: Creating Translation Resources for Southern Uzbek}, 
      author={Mukhammadsaid Mamasaidov and Azizullah Aral and Abror Shopulatov and Mironshoh Inomjonov},
      year={2025},
      eprint={2508.14586},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2508.14586}, 
}

Contacts

We believe that this work will enable and inspire all enthusiasts around the world to open the hidden beauty of low-resource languages, in particular Southern Uzbek.

For further development and issues about the dataset, please use [email protected] or [email protected] to contact.