README.md

metadata

license: cc-by-4.0

Nuclear Terminology Dataset

This repository contains a cleaned subset of the Korean nuclear terminology dataset, sourced from the Korea Public Data Portal. It was used as part of the training corpus for the AtomicGPT project, a domain-adapted large language model for nuclear engineering developed at the Korea Atomic Energy Research Institute (KAERI).

Dataset Details

Language	Korean / English
Domain	Nuclear Engineering
Source	Korea Public Data Portal (data.go.kr)
Original Provider	Korea Hydro & Nuclear Power (KHNP)
Size	2,000 entries
Format	JSON
License	CC-BY 4.0

Data Format

Each entry consists of a nuclear term and its definition, structured as a grammatically natural Korean sentence.

{
  "text": "핵분열은 무거운 원자핵이 중성자를 흡수하여 두 개 이상의 가벼운 원자핵으로 분열되는 현상이다."
}

Preprocessing

The original dataset was obtained from the Korea Public Data Portal and preprocessed as follows:

• Extracted term and definition fields from the original CSV file
• Filtered out entries with empty fields or insufficiently short definitions
• Retained only entries ending with a complete Korean sentence (e.g., "이다.", "된다.")
• Final cleaned subset of 1,000 entries selected for public release

Source

• Korea Public Data Portal: https://www.data.go.kr/data/15038485/fileData.do
• Original Provider: Korea Hydro & Nuclear Power Co., Ltd. (한국수력원자력)

Related Resources

• AtomicGPT Models: KAERI-MLP
• AtomicGPT Evaluation Datasets: KAERI-MLP/atomicgpt-eval-datasets

Citation

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

@article{atomicgpt2026,
  title={AtomicGPT: A Domain-Adapted Large Language Model for Secure On-Premise Applications in Nuclear Engineering},
  journal={Nuclear Engineering and Technology},
  year={2026}
}

About

Developed by the MLP team at the Korea Atomic Energy Research Institute (KAERI).