Skip to content

Dataset · nvidia

Nemotron Korean synthetic personas

nvidia/Nemotron-Personas-Korea

A large set of synthetic personas based on real South Korean demographic distributions.

What's inside

The Nemotron-Personas-Korea dataset contains synthetic personas based on real demographic and geographic distributions of South Korea. Personas have a name, gender, age, marital status, education, profession, and region; data is synthesized based on statistics from KOSIS, the Supreme Court, National Health Insurance Service, KREI, and NAVER Cloud. This is the first major Korean-language persona dataset, CC BY 4.0 license; modalities are image and text, size ranges from 1 to 10 million.

What you can do with it

  • Train a Korean LLM on personalized dialogues
  • Create a LoRA for persona generation
  • Test models for demographic biases
Tasks
Text generation
Languages
Korean
Size
1M–10M rows
Downloads this month
7.3K
Likes
594

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-Personas-Korea")