metadata

license: apache-2.0
configs:
  - config_name: default
    data_files:
      - split: az_en
        path: data/az_en-*
      - split: az_fr
        path: data/az_fr-*
      - split: az_ru
        path: data/az_ru-*
dataset_info:
  features:
    - name: source
      dtype: string
    - name: target
      dtype: string
  splits:
    - name: az_en
      num_bytes: 37429
      num_examples: 100
    - name: az_fr
      num_bytes: 40516
      num_examples: 100
    - name: az_ru
      num_bytes: 41822
      num_examples: 100
  download_size: 77947
  dataset_size: 119767
language:
  - az
  - en
  - fr
  - ru
task_categories:
  - translation
size_categories:
  - n<1K

Azerbaijan Multilingual Translation Benchmark

This dataset provides parallel text data for benchmarking machine translation quality between Azerbaijani (az) and multiple target languages:

Azerbaijani-English (az_en)
Azerbaijani-French (az_fr)
Azerbaijani-Russian (az_ru)

Structure

The dataset contains three splits, one for each language pair:

az_en: Azerbaijani-English translation pairs
az_fr: Azerbaijani-French translation pairs
az_ru: Azerbaijani-Russian translation pairs

Each split has the same structure with two columns:

source: Text in Azerbaijani
target: Text in the target language (English, French, or Russian)

Usage

from datasets import load_dataset

# Load the entire dataset
dataset = load_dataset("LocalDoc/az_multilingual_pairs")

# Access specific language pairs
az_en_data = dataset["az_en"]
az_fr_data = dataset["az_fr"]
az_ru_data = dataset["az_ru"]

# Example of accessing the first entry
print(az_en_data[0])

Purpose

This dataset is designed for benchmarking translation quality between Azerbaijani and multiple languages. It can be used to evaluate machine translation systems, multilingual models, and language-specific capabilities for low-resource language pairs.