Datasets:

dogdoh
/

dari-enko-corpus-sample

dataset_info:
  features:
    - name: source_en
      dtype: string
    - name: target_ko
      dtype: string
    - name: domain
      dtype: string
    - name: quality_score
      dtype: float32
    - name: labse_score
      dtype: float32
    - name: comet_score
      dtype: float32
license: cc-by-4.0
language:
  - en
  - ko
size_categories:
  - 1K<n<10K
tags:
  - translation
  - parallel-corpus
  - en-ko
  - technical-translation
  - mtpe

Dari EN↔KO Technical Translation Corpus (Sample)

🌉 Dari (다리, "bridge") — 67M+ EN↔KO parallel sentence pairs for technical translation.

Dataset Description

This is a 10,000-pair sample from the full Dari corpus. The full corpus contains 67.2 million high-quality EN↔KO parallel segments across multiple technical domains.

Domains

Domain	Full Corpus Pairs
Patent (KIPRIS)	15M+
Medical (PubMed)	12M+
IT/Software	10M+
Legal	8M+
General	22M+

Files

data/sample.csv — 10,000 parallel EN↔KO sentence pairs (stratified by domain)
data/glossary.csv — 5,000 bilingual term entries
data/stats.csv — Domain distribution statistics

Quality Metrics

LaBSE semantic similarity scores
COMET translation quality scores
Human quality ratings (0-1 scale)

Usage

from datasets import load_dataset
ds = load_dataset("dogdoh/dari-enko-corpus-sample")

Full Corpus Access

Tier	Price	Includes
Free	$0	This 10K sample
API Basic	$49/mo	50K chars/month, REST API
API Pro	$199/mo	500K chars/month, batch export
Enterprise	Custom	Full corpus license, on-prem

🔗 API: dari.is-a.dev 📧 Contact: dogdoh1338@icloud.com