The Georgian-English News Parallel Corpus is a curated dataset consisting of 639 professionally aligned article pairs. This dataset bridges the gap between high-quality English tech journalism and natural Georgian translations, focusing specifically on artificial intelligence, enterprise software, and global venture capital.
🚀 Quick Start
from datasets import load_dataset
# Load the parallel corpus
dataset = load_dataset("iraklixyz/georgian-english-news-parallel-corpus", "data", split="train")
# Access the first row
print(dataset[0]["title_ka"])
📊 Dataset Summary
- 📈 Records: 639
- 📋 Columns: 5
- 🗣️ Languages: Georgian (
ka), English (en) - 🎯 Primary Use: Model alignment, translation fine-tuning, and seeding synthetic instruction datasets.
📋 Schema
| Column | Type | Description |
|---|---|---|
title |
string |
Original headline in English. |
title_ka |
string |
Aligned headline in Georgian. |
content |
string |
Full article body in English. |
content_ka |
string |
Full article body translated into natural Georgian. |
meta_description |
string |
A one-sentence Georgian summary of the article. |
⚙️ Curation Notes
Unlike typical machine-translated datasets, this corpus focuses on preserving the technical nuances of the source material. It is particularly useful for:
- Terminology Mapping: Standardizing how concepts like "Inference Computing" or "Open-Source Agents" are expressed in Georgian.
- Knowledge Distillation: Using the English content to verify the factual grounding of the Georgian translations.
- Style Transfer: Training models to adopt a journalistic and objective tone in Georgian.
📋 Statistics
| Metric | Value |
|---|---|
| Total Rows | 639 |
| Unique Content (%) | 100% |
| Domain | Technology & Business News |
- Downloads last month
- 22