DatarrX/PlanetFacts-QA-MY

📝 PlanetFacts-QA-MY

PlanetFacts-QA-MY is a high-quality, human-curated Burmese Question-Answering (QA) dataset focused on planetary science and astronomy. The dataset features localized information derived from official NASA articles, transformed into a natural, conversational Burmese format.

This dataset is specifically built to address the low-resource challenges of the Burmese language in Natural Language Processing (NLP) and to train future language models to understand and reply to scientific queries in polite, natural Burmese speech.

Dataset Description

• —Repository Maintained by: DatarrX
• —Author & Lead Developer: Khant Sint Heinn (Kalix Louis)
• —Language: Burmese (Myanmar) - 100% Unicode Conforming
• —Task: Closed-Domain Question Answering (Science & Astronomy)
• —Format: JSON Lines (.jsonl)
• —License: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)

Dataset Structure

Data Format

The dataset is provided in standard .jsonl format. Each row consists of a single dictionary containing four key fields:

Data Instance Example

{
  "question": "နေအဖွဲ့အစည်းထဲမှာ ဘယ်ဂြိုဟ်က အသေးဆုံးလဲခင်ဗျာ။",
  "answer": "နေနဲ့အနီးဆုံးမှာရှိတဲ့ ဗုဒ္ဓဟူးဂြိုဟ်က ကျွန်တော်တို့ နေအဖွဲ့အစည်းထဲမှာ အသေးဆုံးဂြိုဟ်ပါပဲ။",
  "source": "[https://science.nasa.gov/mercury/facts/](https://science.nasa.gov/mercury/facts/)",
  "planet": "Mercury"
}

Dataset Statistics

• —Current Number of Rows: 399 rows (🚧 ongoing / active development 🚧)

Linguistic Characteristics & Quality Control

1. 1.Diglossia & Tone Strategy The Burmese language exhibits a strong linguistic phenomenon known as Diglossia, consisting of a formal literary style (စာပေဟန်) and a spoken colloquial style (စကားပြောဟန်).

• —📑 Applied Style: This dataset intentionally utilizes the Spoken/Colloquial Style (စကားပြောဟန် – အနိမ့်) to mimic how real-world users interact with AI assistants, voice bots, and search engines.
• —🗣 Tone: Every instance is curated using a Polite and Respectful Tone (e.g., utilizing appropriate sentence-ending particles like "ခင်ဗျာ" or "ပါပဲ") to ensure conversational AI models trained on this data remain polite and user-friendly.

1. 1.Orthographic Accuracy To prevent dataset noise and ensure high-quality text representation, all text entries have been rigorously verified against the Myanmar Orthography Dictionary, 2003 (မြန်မာစာလုံးပေါင်း သတ်ပုံကျမ်း၊ ၂၀၀၃). There are zero font encoding errors; all texts are natively encoded in standard Burmese Unicode.

1. 1.Factual Verification Every question-answer pair is backed by factual documents from NASA’s Planetary Science Division. The inclusion of the source field guarantees transparent verifiability, enabling researchers to use this for Retrieval-Augmented Generation (RAG) evaluation.

Intended Uses & Limitations

Intended Uses

• —Fine-tuning large language models (LLMs) for Burmese conversational tasks.
• —Evaluation dataset for RAG systems to measure factual retrieval against Burmese queries.
• —Researching low-resource domain localization (converting English scientific tables into natural Burmese Q&As).

Limitations

• —The dataset is currently under active extension (currently at 399 rows).
• —It covers only topics related to our Solar System based on NASA resources.

Licensing & Attribution

This dataset is released under the CC BY-SA 4.0 license.

• —You are free to: Share, adapt, and use this dataset for both non-commercial and commercial applications.
• —Under the following terms: You must give appropriate credit by citing the author and DatarrX foundation, and provide a link to the license. If you remix, transform, or build upon the material, you must distribute your contributions under the same license.

📜 Citation & Academic Reference

If you incorporate the PlanetFacts-QA-MY dataset into your academic research, machine learning experiments, language modeling benchmarks, or Retrieval-Augmented Generation (RAG) evaluation pipelines, please acknowledge the author and publisher via the following official BibTeX citation:

@misc{datarrx_planetfacts_2026,
  author       = {Khant Sint Heinn},
  title        = {PlanetFacts-QA-MY: A Human-Curated Polite Spoken-Style Burmese Question-Answering Dataset Based on NASA Planetary Science Resources},
  year         = {2026},
  publisher    = {Hugging Face},
  organization = {DatarrX},
  howpublished = {https://huggingface.co/datasets/DatarrX/PlanetFacts-QA-MY},
  note         = {Published under DatarrX. Open-source community asset released under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)}
}

About the Author & Organization

Lead Curator: Khant Sint Heinn (Kalix Louis)

Khant Sint Heinn (Burmese: ခန့်ဆင့်ဟိဏ်း) is a Machine Learning Engineer specializing in Natural Language Processing (NLP), data foundations, and open-source AI development. His work focuses on strengthening the Burmese (Myanmar) tech ecosystem by building scalable, clean datasets and critical language infrastructure to transition Burmese from a low-resource language into an AI-ready asset.

• —Profiles: GitHub | Hugging Face | Kaggle

Organization: DatarrX

DatarrX (Burmese: ဒေတာ-အက်စ်) is a non-profit, community-driven open-source foundation dedicated to building a robust digital foundation for the Burmese language in the AI era. DatarrX works to democratize technical resources, localize documentations, and build high-quality AI benchmarks.

• —Profiles: GitHub | Hugging Face | Kaggle